Note: Descriptions are shown in the official language in which they were submitted.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
M>isTHOD AND SYSTEM TO CONSTRi~CT ACTION COORDINATION
PROFILES
BACKGROUND OF THE INVENTION
1.1. Technical Field
This invention is a method or system to construct at least one profile
representing
how the actions of an object of investigation are coordinated, the profiles)
being
based on computed measures of longitudinal association or temporal contingency
that quantify patterns o~ interaction in repeated measures or time series data
that
include two or more variables for one individual.
1.2. Description of Related Art
Scientific knowledge often is represented in the form of mathematical models.
Prior art related to this invention will be described in the context o~
computational
z0 methods and systems to create, verify, and refine models that represent
objects of
investigation.
The statistical method is a primacy computational method to inform the process
of
model building. This inventian addresses certain fundamental problems that
derive
from limitations often encountered when the statistical method is a primary
means
to inform the process of creating, verifying and reFining mathematical models.
By
addressing these problems, this invention facilitates many scientific
investigations
and practical arts that may benefit from scientific knowledge.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
The Appendix is an outline that helps reveal the logical structure of this
application.
1.2.1. The Need to Measure Interactions that ai°e Temporal
Contingencies
Model builders generally identify an object to model and abstract variables
that
may be relevant to its functioning. Then modelers determine how the variables
interact in order to inform the process of model construction. To a large
extent,
mathematical models are veriFed by the extent to which they accurately
represent
1 o interactions of their obj ects in the real world.
The statistical method is an important tool for constructing many mafhematical
models. The statistical method includes various measures and procedures for
revealing interactions that can be modeled.
A primary problem addressed by this invention derives from the fact that the
statistical method is best suited to address objects of investigation that are
collective entities. Groups, samples, and populations axe collective entities.
Section 1.2 of parent patent application 091170,956 describes many limitations
and
problems of the prior art. Many of these problems, limitations, and solutions
are
illustrated in the context of clinical trials. This invention also addresses
these
problems and limitations but generally in the broader context of complex
systems.
Section 1.2.2 of this document emphasizes problems more specifically addressed
by this invention. Section 2.~. of this document describes how this and the
parent
invention address the following problems.
1.2.2. Speci Fic Problems Involved in fine Prior Art
3d The statistical method and mathematical models often are used to
investigate
complex systems. For example, mathematical models based on statistical
analyses
of group data have been used to model the apparent effects of cholesterol and
other
lipid fractions on mortality and major cardiovascular health events. Such
models
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
serve important functions. For example, mathematical models have used
laboratory
data to predict the long-term health effects of new cholesterol lowering
drugs.
Nevertheless, statistical analyses have important limitations for
investigations of
complex systems.
Conventional applications of the statistical method are best suited for
analyses of
cross-sectional data for collective entities and for predicting events such as
death
That are not recurrent for individuals. Statistical analyses axe not as well
suited to
measure longitudinal associations or temporal contingencies between and among
1o variables within individuals - interactions that become evident in
longitudinal,
repeated measures, or time series data.
The statistical method does have some functionality for analyzing repeated
treasures data, especially far groups. For example, the statistical method
often is
used to analyze change scores such as pre-post differences in clinical trials.
However, this functionality becomes limited as the number of repeated
measurements increases. This limitation is due to the fact that the number of
differences between any two measurements increases rapidly with the number of
repeated measurements. In addition, it is not meaningful, appropriate or
useful to
2o conduct statistical tests on all differences that are possible when there
are more
than a few repeated measurements.
The statistical method also includes techniques such as repeated measures
analysis
of variance. However, the usefulness of such techniques tends to be limited
when
the levels of one or more independent variables differ across many repeated
measurements for each individual. For example, generally it is not feasible
with
conventional analyses to substitute blood levels of drug for planned doses
before
rerunning analyses of the effects of treatment on health.
Conventional data analysis procedures are of limited value for supporting
detailed
yet comprehensive investigations of complex individual systems whose variables
may interact in a nonlinear manner. Here is additional information about five
problem areas that are mentioned in the preceding statement - individuality,
3
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
complexity, nonlinearity, comprehensiveness, and detail - together with a
statement about the need to address all these problem areas as a set in
particular
investigations.
1.2.2.1. Problems Involving Individuality
The statistical method is best suited for analyses of cross-sectional data for
collective entities such as groups. Many statistical descriptions and
inferences are
about measures of central tendency for groups. Statistical analyses often are
based
on assigmnents of individuals to groups such as treated or not treated,
responder or
non-responder. The results of such analyses apply most directly to collective
entities.
The fundamental limitation of the statistical method that involves
individuality will
be viewed from two perspectives: (1) the application and (2) the discovery of
scientific knowledge. Both perspectives will be illustrated by example.
The statistical method often is applied to describe groups and to use sample
data to
make inferences about populations. Statistical inferences are used to draw
generalized conclusions and make predictions. The extent to which generalized
conclusions and predictions about collective entities apply to individuals
generally
is limited. This can be illustrated in the context of group clinical trials.
Individual
patients axe not apt to experience the same safety and efficacy as the average
patient in a clinical trial.
The extent to which generalized conclusions and predictions about populations
apply to individuals depends on the extent fo which individuals are typical of
groups. It also depends on the extent to which samples represent populations -
at
least with respect to all considerations relevant to particular investigations
- as
well as how members of samples are assigned to treatment groups.
Science is accounting for more and more factors that affect the responses of
patients to medical treatments. For example, advances in genetics are
identifying
4
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
many ways in which individuals differ in manners that are relevant to disease
and
response to treatment. People need better ways to individualize treatment.
The fundamental limitation of the statistical method with respect to
individuality
also has profound implications for scientific discovery. This will be
illustrated in
the context of functional genomics and proteomics as it involves health
disorders
and medical treatments.
Now that genomes are being mapped, some high priority tasks are to identify
how
the products of gene expression function together and to identify how genetic
differences that distinguish individuals, such as single nucleotide
polymorhpisms,
affect biological functions and responses to treatments. Such tasks currently
are
hampered by a lack of methods that can be applied to individuals to measure
how
proteins interact to control biological functions, of how treatments affect
protein
interactions, and of how treatments interact with proteins and health
variables.
Measurement of such interactions for individuals, as distinct from groups, is
becoming increasingly valuable as it becomes easier to identify how
individuals
differ genetically.
2o Group assignments and averages tend to obscure effects of genetic
differences on
health for individuals and their individual responses to treatments. This
makes it
difficult to identify genetic differences and form classifications that are
predictive
of health disorders and differential responses to drugs. This in turn makes it
difficult to target drugs to the right pafients during drug development and
during
clinical care.
1.2.2.2. Problems Involving Complexity
Complexity derives from the fact that individual systems often have many
parts,
have different types of action, and function in various and chan grog
environments.
Furthermore, certain concepts that often are applied to individuals have
various
manifestations. For example, health of persons is manifested at different
levels of
measurement hierarchies such as through laboratory measures, signs and
5
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
symptoms of disorder, measures of physical and mental fimcfioning, and
measures
of quality of life.
Complexity in the context of empirical investigations often becomes eviden t
by the
S fact that many variables are available to describe individuals and their
environments. Furthermore, many of these variables interact in various
combinations. Investigators and practitioners need better methods and systems
to
quantify, discover, and describe many interactions simultaneously.
I.2.2.3. Problems hlvolving ~lonlinearity
Interactions between and among variables that describe complex systems often
are
not linear. Two aspects of nonlinearity can be illustrated in the context of
multiple
linear regression, a commonly used statistical procedure for creating
mathematical
15 models. Multiple linear regressian models describe the functional
relationship
between a dependent variable, y, and a set of dependent variables, x1, x~,..
.x".
Two aspects of linearity, proportionality and additivity, can be illustrated
with the
equation y = 4 + Sx~ + 2x2. Fox this equation, Each one-unit increase in xj
yields a
20 5-unit increase in y regardless of the value of x1. This illustrates
proportionality.
Furthermore, the effects of x, and x~ in this equation are additive. T3owever,
complex systems often manifest nonlinear interactions. People need improved
methods and systems to address nonlinearity.
25 1.2.2.x.. Problems Involving Comprehensiveness
A praductive but conventional experimental research strategy is to isolate
independent variables and investigate their effects one by one. Such research
often
is hypothesis driven - hypotheses that may be rejected by statistical tests
based on
3~ group data for collective entities. This isolate-and-test strategy fends
toward
simplified models that do not reveal haw many variables, parts, and
manifestations
of complex systems interact in coordinated manners,
G
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
The failure to measure how complex systems interact in coordinated manners is
a
problem because coordinated action is a hallmark of how complex systems
function in interesting and important ways. We need improved methods and
systems to investigate how variables, parts, and manifestations of complex
systems
function together to regulate and sustain themselves as whole individuals that
act
as agents and respond. Such methods and systems would be more comprehensive
of how many variables, parts, and manifestations of individual complex systems
interact.
Biology is beginning to recognize the limits of the isolate-and-test strategy.
Dr.
Leroy Hood and the Institute for Systems Biology advocate systems biology
(http:l/www.systemsbiology.org/workwhat.html). They explicitly recognize that
one cannot learn about biological systems by studying one gene or protein at a
time. They recognize the need to study interactions within and across levels
of
biological information. They recognize that complex systems give rise to
emergent
or systems properties such as abilities of brains to learn and remember.
Dr. Hood has described this new approach to biology as "discovery science." He
contributed to the initiation of the Human Genome Project - a prime example of
discovery science. "Discovery science enumerates the components of particular
objects independent of the questions that characterize the hypothesis-driven
science commonly practiced in biology"
(http://www.systemsbiology.org/workhist.html).
A recent article on the yeast galactose-utilization pathway was considered by
the
authors to demonstrate "proof o~ principle" of the systems approach to biology
(T.
Ideker, V. Thorsson, J.A. Ranish, R. Christmas, J. Buhler, J. K. King, R.
Bumgarner, D. R. Goodlett, R. Aebersold, L Hood, Science, 292, 929-934, 2001
).
Although the objective of this research was "to build, test, and refine a
model of a
cellular pathway" using, among other things, information about protein-protein
interactions, there appears to be no global or comprehensive aftempt of
measure
the interactions using time series data on protein levels,
7
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Although the need for comprehensive methods and systems for measuring
interactions has been illustrated in the context of biology, similar problems
plague
investigations of many other types of complex system.
1.2.2.5. Problems Involving Detail
The need for detailed investigations becomes evident in at least two different
ways.
First, it often would be valuable to investigate many different variables in
particular investigations. This can be illustrated with the rating scales that
often are
to used in clinical trials for antidepressant drugs. Such composite rating
scales often
include many items measuring different things such as mood, movement,
ideation,
and sleep. There is need for more effective methods to investigate the effects
of
drugs both across all items and for detailed investigations of drug effects on
individual items.
Second, there is need for more detailed investigations with respect to each of
fine
individual variables that may be investigated, for example, in clinical
trials. For
example, it may not be enough to investigate how a particular dose of drug
affects
depression. There also is need to investigate treatment effects as functions
of dose
ox blood levels of drug, episodes of treatment, as well as delay and
persistence of
response to treatment - both for individual patients and for groups of
patients.
1.2.2.6. Need to Investigate All Five Types ofProblem as a Set
Various techniques have been developed to address at least some of the
problems
just described. However, the prior art tends to address the particular
problems
individually. This piecemeal approach does not recognize that all five types
of
problem are of one cloth. All five types of problem need to be addressed as a
set.
Tradeoffs between, for example, detail and comprehensiveness for particular
3o investigations should not be forced by the limitations of methods and
systems used
to process data.
8
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Important aspects of the dynamic involving different strategies o~ scientific
investigation can be discussed in terms of problems in this set. One example
is the
dynamic between comprehensive and detailed investigations. Distinctions among
the sciences themselves such as chemistry, biology, and psychology can be
viewed
as attempts to limit the comprehensiveness of investigations. A fundamental
and
productive research strategy is to focus particular eFforts on ever more
detailed
investigations of ever more delimited sets of phenomena. On the other hand,
many
people recognize the need to investigate camplex wholes. We need better
methods
and systems to accommodate both strategies simultaneously.
1.2.3. Citations
U. S. Patent No. 6,055,491 involves a method and apparatus for analyzing co-
evolving time sequences.
U. S. Patent No. 6,249,755 and U. S. Patent No. 5,528,516 involve an apparatus
and method for event correlation and problem reporting.
U. S. Patent No. 6,173,240 presents multidimensional uncertainty analysis.
U. S. Patent No. 6,134,510 describes a method for detecting synch ronicity
between
several digital measurement series with the aid of a computer.
U. S. Patent No. 6,098,024 addresses a system for process data association
using
LaPlace Everett interpolation.
U. S. Paten t No. 6,051,209 covers a method of evaluating the effects of
administering external stimuli or a treatment on the brain using positron
emission
tomography.
3Q
Section 1.2.2.4 cites two web pages. Sectian 4.2.4 also quotes the Iirst of
these two
web pages. The two web pages are:
9
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Institute for Systems Biology, What is Systems Biology, URL
http:llwww.systemsbiology.org/workwhat.html, Viewed 4/1001; and
Institute for Systems Biology, History of Concepts Leading to the Institute,
URL http:/lwww.systemsbiology.org/workhist.html, Viewed 4/10/01.
Section 1.2.2.4 also cites the following article:
T. Ideker, V. Thorsson, J.A. Ranish, R. Christmas, J. Buhler, J. K. King, R.
Bumgarner, D. R. Goodlett, R. Aebersold, L Hood, Science, 292, 929-934,
2001.
Data for the hormone data example in Section 4.9 were described and presented
in
the following citations:
Padmanabhan, V., McFadden, K., Mauger, D.T., Karsch, F.J., and Midgley,
A.R. (1997). Neuroendocrin a control of follicle-stimulating hormone (FSH)
secretion. 1. Direct evidence for separate episodic and basal components of
FSH secretion. Enclocrifaology 138, 424-432, and;
Midgley, A.R., McFadden, K., Ghazzi, M., Karsch, F.T., Brown, M.R.,
Mauger, D.T., and Padmanabhan, V. (1997). Nonclassical secretory
dynamics of LH revealed by hypothalamo-hypophyseal portal sampling of
sheep. Etadocrir2e 6, 133-143.
BRIEF SUMMARY OF THE 1NVENTIO
This invention is a method or system to constniot at least one profile
representing
how the actions of an object of investigation are coardinated, the profiles)
being
based on computed measures of longitudinal associafion or temporal contingency
that quantify patterns of interaction in repeated measures or time series data
that
include two or more variables for one individual. Such profiles are called
action
coordination proFles (AGPs).
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
AGPs can provide quantitative descriptions of how individual complex systems
may control and regulate themselves, of how two or more individual systems may
interact, of how complex systems may be controlled or affected by their
environments including treatments, and of how individual systems may control
or
affect their environments.
In practice, AGPs are limited to selected variables and episodes of action for
particular objects of investigation. This is illustrated by the examples in
Section
~.9. One example involves certain pituitary and reproductive hormones measured
every 5 minutes for up to about 12 hours for individual ewes. Another example
involves variables considered to affect the Gross Domestic Product of the
United
States economy using quarterly data for about ~.2 years.
The title of parent application Serial No. 091470,956 is "computational Method
and System to Perform Empirical Induction." Empirical induction involves
procedures to draw generalized conclusions and make predictions from data.
More
specifically, this invention and its parent involve computational procedures
to
provide high quality generalized conclusions and predictions as high quality
is
defined in Section 1.2 of the parent application.
The key innovative concept for this invention and its parent comprises a
computational method and system specifically designed to process repeated
measures and time series data to rr2easz~r~e interactions between and among
variables fox objects of investigation that are individuals. The parent
application
describes the Method for the Quantitative Analysis of Longitudinal
Associations
(MQALA~.
MQALA and AGPs include an extensive set of computational tools and analytic
options that users can select and apply to address many types of problem
encountered in scientific investigations and practical affairs. All these
tools and
analytic options are based on a core set of computational methods or systems.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
This invention and its parent are distinct from and often complementary to the
statistical method. As such, this invention facilitates scientific
investigations of
individuals both as individuals and as members of collective entities. For
example,
these inventions often can be used to facilitate both the individualization of
medical care and the conduct of group clinical trials for treatments used to
control
or manage chronic disorders.
Since AGPs are a direct extension and distinct improvement on the parent
application, much material in the parent application also applies to ACPs.
Many
terms used in this application are defined in Section 2.9 of the parent
application.
The following subsections provide a brief summary of the structure of ACPs as
well as how they are constructed, functions of ACPs, and how ACPs address
limitations of the statistical method as well as the five specific previously
identified problems involved in the prior art.
2.1. Structure ofACPs
An ACP can be characterized as a set of computed measure values, the set
having
two dimensions. One dimension represents independent events and a second
dimension represents dependent events. Each column or row for the dimension
representing independent events corresponds to one of two or more variables or
sets of variables or the results of applying certain features used to define
independent events. Each column or row for the dimension representing
dependent
events corresponds to one of two or more variables or sets of variables or the
results of applying certain features used to define dependent events.
Table 1 illustrates the general structure of an ACP with 10 variables and only
one
column or row for each variable. The same variables are used for both
dimensions.
Rows represent the variables functioning as independent variables (IVs) to
define
independent events. Columns represent the variables functioning as dependent
variables (DVs) to define dependent events. Cells are forn~ed at intersections
of
rows and columns.
12
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Table 1 uses the symbols "o" and "*" to represent scores or measure values in
general. Each cell of Table 1 that contains an "*" indicates that the score or
measure value was obtained when the variable labeling a row was functioning to
define independent events. Each cell of Table 1 that contains an "o" indicates
that
the score or measure value was obtained when the variable labeling a column
was
functioning to define dependent events. There are no measure values for cells
on
the concordant diagonal, which are represented with the symbol "-". Additional
columns and rows would be used to represent Boolean events defined on two or
more variables, to represent transition events, or to represent additional
ways of
defining independent or dependent events.
Table 1. Structure of an ACP with 10 variables.
Des
1 ' ~ i 5 6 ~ 8 ~ ~
2 3 4 7 9 10
i ~l - ~ o 0 0 0 0 0 0 o
o
2 * - o o ' i ' ~ o ' 0
o o o 0
3 * * ~ ~ 0 o I o i 0
_ 0 o o
I '~ ~ * ~ I I o o ~ o ~ '
* * _ o o o
LVs $ * * * * _ o 0 0 0 0
6 * * * * * - o 0 0 0
i ~ I
* * I * ~ * _ o ' 0
* * 0
~g ~ * * * * * * ~ _ ~ 0
* 0
9 * ~ * * * * * ~ * _ ~
* o
* * *
Each cell in Table 1 that is identified by an "*" or an "o" represents a
particular
interaction. Particular interactions also can have dimensions. Dimensions for
particular interactions represent analysis parameters such as level of the
independent variable, level of the dependent variable, delay, persistence,
episode
length and episode criteria for the independent variable, and episode length
and
episode criteria for the dependent variable.
The term "dimensian" is being used in two contexts. In the context of ACPs,
"dimension" refers to variables functianing to define either independent or
13
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
dependent events. In the context of particular interactions that are
represented by
cells, "dimension" refers to analysis parameters that may or may not have
multiple
levels.
The computed measure values in ACPs generally are either longitudinal
association scores or values of measures derived from longitudinal association
scores. Section ~.4 identifies examples of measures that can be used to
construct
ACPs. Typically, the magnitude of each score or measure value in ACPs
quantifies
either the amount of evidence for a longitudinal association or the strength
of that
l0 association. The signs of longitudinal association scores indicate positive
or
negative associations. Positive scores indicate that dependent events are more
apt
to occur in the presence of independent events than in the absence of
independent
events. Negative scores indicate that dependent events are less apt to occur
in the
presence of independent events than in the absence of independent events. Zero
scores or measure values indicate no evidence for longitudinal associations or
temporal contingencies.
Each of the variables used to define independent and dependent events for ACPs
would need to be measured or assessed repeatedly for an individual on two or
more
occasions. In addition, each variable should have the potential to vary -
fluctuate in
level or recur over time - for the object of investigation represented by the
ACP.
Variables could be transformed mathematically before computing scores or
measure values in ACPs.
ACPs can be portrayed as tables, figures, graphs, and displays. It is
recommended
that columns and rows in ACPs for particular fiypes of investigation be
presented in
standardized orders to facilitate comparisons and analyses of profiles for
different
individuals or for different episodes of action.
Unless otherwise specified, fh a same variables and features would be used in
the
same way to define both independent and dependent events for ACPs. This means
that people who construct ACPs generally need not identify variables or evenfs
as
independent or dependent. Furtllernore, this is in accord with how complex
1 ~1
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
systems often function. A given event may function as a dependent event with
respect to some other events and the same given event may function as an
independent event with respect to still other events.
In some cases, events may function in feedback loops to affect mare events of
the
same type. For example, neurotransmitters can have both pre- and post-synaptic
receptors so that release of a transmitter can help propagate a signal and
feed back
fo affect release of additional transmitter.
Features of MQALA can be used alone or together with experimental procedures
to help distinguish causal from non-causal associations. Some portions of ACPs
could remain blank if, for example, investigators determine that it would not
be
meaningful to consider variables that were under control in experimental
investigations to function as dependent variables.
1s
Typically, various analysis parameters would be used to obtain the measure
values
in ACPs. Level of independent variable and level of dependent variable are
required analysis parameters when the variables are dimensional (when a series
of
values for a variable has more than two different values) and the user of
MQALA
decides to examine more than two levels.
Another analysis parameter, delay, would be a primary analysis parameter when
ACPs are used to investigate the temporal criterion of causal and other
predictive
relationships (Section 1.8.9). Delay is defined on variables functioning as
independent variables. Users of ACPs could specify one or more particular
values
of delay or a range of values. One ACP or portion of an ACP would be computed
for each particular value of delay. In addition or alternatively, one ACP
could
summarize scores across a range of values of delay.
Additional analysis parameters for interactions that are described in the
parent
patent application include episode length and episode criterion for
independent and
dependent variables as well as persistence defined on variables functioning as
IVs.
users can define additional analysis parameters. Typically, the same scoring
IS
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
options would be selected for each IV and for each DV that are used in
particular
ACPs.
Typically, scores actually shown in ACPS would be summary scores. Information
about the location of each summary longifudinal association score or
derivative
measure in an array identifies the conditions that yielded the summary
measure.
These conditions are defined in terms of features such as analysis parameter
levels
and Boolean events. Users would be able to drill down from scores shown in
summary ACPs to examine scores as functions of analysis parameter levels and
in
terms of Boolean events.
2.2. Functions of ACPs
ACPs are a way of displaying particular types of quantitative information so
that it
can be used to discover and describe patterns of longitudinal association or
temporal contingency between and among variables and events. Use of ACPs to
discover and describe patterns in a systematic, comprehensive, and detailed
manner will advance the objectives of scientific investigation, the conduct of
practical affairs, and decision-making.
The author has coined various terms to describe AGPs and the methodology upon
which they are based. These terms emphasize different ways in which the
technology is unique and of value.
MQALA can be viewed as a contribution to "temporal contingency analysis." The
contingencies (associations) involve independent and dependent events dei=Lned
in
multidimensional spaces formed primarily by applying analysis parameters and
Boolean operators to transformations of repeated measures data including time
series.
Independent and dependent events can be defined in great detail. Analysis
parameters account for things such as levels of independent (predictor) and
dependent (predicted) variables. Optional analysis parameters account for
things
1G
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
such as episodes of events. Here is an example of an independent event defined
using such parameters and a temporal resolution of one day. Did or did not a
given
patient meet the criterion on each of a series of days of taking 100 mg or
more of a
given drug (Dnig 1) on 5 out of 7 cansecutive days? Additional optional
analysis
parameters can be used to define temporal aspects (delays and persislen cies)
of
relations between events.
Boolean operators can be applied to events defined with analysis parameters to
define additional events called Boolean events that are based on two or more
to independent or two or more dependent variables. For example, a Boolean
independent event could consist of meeting the criterion defined for Drug 1 in
the
preceding paragraph AND the criterion of taking Drug 2 at a dose of 50 ing or
more on 4 out of 6 consecutive days. The presence of such a Boolean AND
independent event may be sufficient, for example, to increase the presence of
a
particular type of dependent event such as the level of a liver enzyme being
above
the upper limi of normal.
MQALA analyzes such contingencies between independent and dependent events.
Many thousands of different events and types of events can be analyzed
simultaneously in particular investigations to identify the levels of analysis
parameters and the Boolean events that yield the most evidence for
associations or
interactions.
The "temporal" in "temporal contingency analysis" indicates that MQALA and
any particular ACP quantifies and describes the directions and amount of
evidence
for contingencies (associations or interactions) between and among events as
these
contingencies are evident in data that are about the individual and are
collecfed
over time from the individual. MQALA also quantifies the sfrength of
associations,
contingencies, or interactions. MQALA's capability to analyze temporal
contingencies derives from the Fact that it is applied to longitudinal,
repeated
measures, or time-series data as relatively distinct from cross-sectional
data.
17
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
ACPs also can be described with coined terms such as "action coordination
f ngerprints," "movement coordination ~ngelprints," "behavior coordination
fingerprints," and "interaction fingerprints." The term "fingerprints" in such
descriptions focuses attention on the fact that AGPs can describe that that is
characteristic of individuals that may be unique or different from other
individuals.
In addition, AGPs can be used to describe that that is characteristic of
episodes of
coordinated action, movement, or behavior far individuals. For example,
episodes
of coordinated locomotion of horses have been characterized as walls, cantor,
trot,
and gallop.
The term "fingerprints" in its conventional use refers to the form or
structure of
skin on the fingers. In contrast, ACPs fingerprint something that is more
abstract
and conceptual - namely the way actions interact. Interactions indicate
coordination. AGPs can fingerprint how individuals function, control, and
sustain
themselves as well as interact with each other and their environments.
2.3. How do AGPs Help Address Limitations of the Statistical Method?
Section 1.2 of this application describes related art in the context of
creating,
verifying, and refining mathematical models that represent objects in the
world.
More specifically, the referenced section presents certain limitations and
problems
related to using the statistical method for this purpose. This section and its
subsections describe how MQALA and ACPs help address these limitations and
problems.
Both MQALA, which now includes ACPs, and the statistical method are distinct
and often complementary computational methods of empirical induction.
Computational methods and systems of empirical induction are used to draw
generalized conclusions and make predictions from data.
Although both MQALA and the statistical method are computational methods of
empirical induction, they are distinct in other key respects. These
distinctions
include the type oFdata (evidence) that the two methods are best suited to
analyze,
18
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
the objectives of analyses, the computational procedures themselves, and the
type
of entities about which conclusions are drawn and predictions are made.
2.3.1. MQALA and the Statistical Method Are Best Suited to Analyze
Distinct Types ofData (Evidence)
MQALA analyzes repeated measures or time series data for particular
individuals.
MQALA requires data for at least two variables or types of events. At least
one
variable must function as an independent variable and at least one variable
must
function as a dependent variable. Both independent and dependent variables
must
vary within individuals in order to obtain nonzero longitudinal association or
benefitlharm scores.
In contrast to MQALA, the statistical method is best suited to analyze cross-
sectional data for groups of individuals. Inferential statistical procedures
(as
contrasted to descriptive statistical procedures) also generally require data
far
independent and dependent variables from groups with two or more individuals
per
group.
Thus, from what has been said about the type of data best suited for analysis
by
MQALA and the statistical method, the two methods generally rely on different
types of evidence for relationships between and among variables. MQALA relies
on longitudinal associations (temporal contingencies) between and among
variables within individuals. In contrast, the statistical method is best
suited to
analyze cross-sectional associations - differences between and among
individuals
or groups of individuals.
The statistical method is best suited for analyses involving groups of
different
individuals at one or only a few times. In contrast, MQALA and AGPs are best
suited for analyses involving one individual at many different times.
19
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
2.3.2. MQALA and the Statistical Method Have Distinct Objectives
Objectives of analyses conducted with MQALA are to quantify, discover, analyze
and describe longitudinal associations (temporal contingencies) between and
among variables within individuals. MQALA provides generalized conclusions
about longitudinal associations between and among variables for individuals.
Such
conclusions are generalized over repeated measurements. MQALA does this with a
variety of scores including scores presented in the form of ACPs.
MQALA also supports predictions. These predictions are about how individuals
will function or respond in the future. Predictions are based on the
assumption that
past experience can be used to help predict the future.
MQALA supports predictions in at least two related ways. First, generalized
conclusions about how an individual has functioned or responded to date can be
used to make predictions about how that individual will respond or function in
the
future. For example, assume that a benefitlharm score based on many repeated
measurements of drug dose and blood pressure for a particular patient over the
course of the last year indicates that the drug had a substantial beneficial
effect for
that patient. This score would support the prediction that the same drug would
continue to have the same beneficial effect for the same patient over the
course of
the next month.
MQALA also supports predictions with a feature called predictive indices.
Predictive indices are one way to use information from two or mare predictors
(IVs) or sets of predictors used to define Boolean events to make predictions
about
a predicted variable (DV). Predictive indices are computed directly from
information used to compute particular longitudinal association scores.
MQALAsupports direct predicfions. That is, the predictions are for fihe same
individual that the data are about. Furtbernlore, the predictions are for fine
same
variables analyzed with the same analytic options.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
MQALA does not directly support inferences from one individual or group of
individuals to another. However, MQALA provides scores and other measures that
can be analyzed statistically to make such inferences when the scores or
measures
are available for two or more individuals. This illustrates the
complementarity of
MQALA and the statistical method.
MQALA is a powerful new set of computational tools for drawing conclusions and
making predictions about individuals by providing quantitative descriptions of
experience that has been recorded as repeated measures data.
In contrast to MQALA, the statistical method is best suited to describe
characteristics of groups and to quantify, discover, analyze and describe
cross-
sectional associations between and among variables for groups of individuals.
In
addition, the statistical method includes procedures for using group
descriptions to
make statistical inferences from samples of individuals to populations of
individuals.
Descriptive statistics axe best suited to describe groups of individuals. The
application of such graup descriptions to individuals is indirect. Similarly,
statistical inferences generally are for groups rather than for individuals.
Conventional parallel group clinical trials are conducted primarily for the
benefit
of groups of patients that may be candidates for freatment in the future. This
fact
often raises ethical questions concerning the patients who actually
participate in
conventional group clinical trials.
2.3.3. MQALA and the Statistical Method Use Distinct ~'omputational
Procedures
The computational procedures for MQALA and the statistical method differ in
several important respects. Unlike the statistical method, MQALA must convert
any dimensional series for independent and dependent variables into sets of
21
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
dichotomous series. All analysis parameters and Boolean events are deFined on
such dichotomous series. Dichotomous series for independent variables and
dichotomous series for dependent variables are cross-classified to yield 2 x ?
tables. This procedure can easily yield thousands of 2 x 2 tables for any
particular
individual.
MQALA continues by computing standardized longitudinal association or
benefitlharm scores for each of these 2 x 2 tables. These scores are
standardized
with respect to all scores that are possible given the marginal frequencies of
observed 2 x 2 tables. Standardization allows the scores to be summarized and
compared. In addition, standardization makes it reasonable to compute overall
benefitlharm scores across many dependent variables for particular
individuals.
Overall bene~tlharm scores can be computed with or without differential
weights.
Longitudinal assoaiatian scores, benefitlharm scores, and overall benefitlharm
scores - one score from each of two or more individuals - can be analyzed
statistically. This illustrates the complementarity of MQALA and the
statistical
method.
2.3.4. MQALA and the Statistical Method Are Best Suited for Distinct
Types of Entities
A key distinction between MQALA and the statistical method involves the type
of
entities for which the methods are best suited to draw conclusions and make
predictions. MQALA draws conclusions and makes predictions about individuals.
For MQALA, individuals include populations investigated as wholes. In
contrast,
the statistical method is best suited to draw generalized conclusions and
support
predictions about groups and populations oFindividuals.
Both MQALA and the statistical method are tools For the conduct of objective
scientific investigations. Systematic scientific knowledge generally is
considered to
involve generalized conclusions rather than particulars.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
MQALA can be used to make generalized conclusions about individuals. Such
conclusions are generalized over time within individuals where time is
represented
by repeated measurements.
In contrast to MQALA, the statistical method is best suited to draw
generalized
conclusions about groups. Descriptive statistics generalize across the
individuals
that comprise groups. Statistical inferences generally are based on group
comparisons and sample data. Such results apply only indirectly to
individuals.
2.~1. MQALA Helps Address Problems Described in Section 1.2.2.
The following subsections provide additional information about each of the
problem areas described in Section 1.2.2 with a focus on the ACP component of
MQALA.
2.1.1. ACPs Help Address Problems Involving Individuality
Since an ACP is computed from data about an individual, the ACP applies most
directly to the individual that the data are about. Section 2.6 of the parent
patent
application discusses differences between direct, indirect, and doubly
indirect
predictions together with some advantages of using direct predictions for
individuals.
Differences between direcf, indirect, and doubly indirect predictions can be
2S illustrated in the context o~ conventional parallel group clinical trials.
Although
such trials provide valuable information about both groups and group members,
the
application of results from such trials to individuals is doubly indirect. One
source
of indireetness involves the extent to which samples represent populations. A
second source of indirectness involves the extent to which particular
individuals
3p are typical of average population members.
The parent application also explains how treasures of longitudinal
association,
such as those used in ACPs, can be reliable and valid measures of longitudinal
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
association for individuals. In brief, applying experimental procedures within
individuals can enhance validity. Collecting and analyzing data from many
repeated measurements can increase reliability.
ACPs can be computed to describe how the parts, variables, and manifestations
of
unique individuals interact. For example, the US economy is a relatively
unique
individual. It generally is not feasible to investigate unique individuals by
sampling
populations and making inferences from the samples. For such reasons, MQALA is
better suited than the statistical method to investigate individuals that are
unique.
ACPs also can be computed for individuals that may be different - not typical -
of
average individuals. For example, patients with high blood pressure can differ
with
respect to concurrent disorders, concomitant treatments, gender, race, age and
other factors. Conventional strategies for investigating treatments favor
homogenous groups of substantial size. It can be difficult to recruit samples
of
substantial size when many factors differentiate patients. The number of
populations that need to be investigated also increases with the number of
factors
that differentiate patients. The number of populations that need to be
investigated
and the number of individuals available in each population clearly limit the
strategy of investigating homogeneous groups. MQALA, including AGPs, address
such problems by providing unique functionality to help enable scientific
investigations of individuals.
Scientific investigations, whether o~ groups or of individuals, have well
known
advantages such as providing abjective and repeatable results. Some unique
advantages of conducting scientific investigations of individuals can be
considered
from the practical and epistemological perspectives.
From a practical perspective, therapy often needs to be individualized because
patients differ from one another in their responses and preferences. MQALA
appears to be the missing key for providing individualized or personalized
health
care that is for people with chronic health concerns and based on objective
scientific procedures for drawing generalized conclusions and making
predictions
?~
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
from data. Chronic disorders and their treatments often are investigated best
with
repeated measures and time series data.
From an epistemological perspective, MQALA can be used to help discover how
individual differences affect susceptibility to disease and response to
therapy.
Differences relevant to both include genetic differences.
Another reason why MQALA is an important analytic tool is that it helps enable
scientific investigations of how individuals interact with their physical and
social
environments. The uniqueness, richness, and continuity of such interactions
appear
to be the essence of individual identity.
2.4.2. ACPs Help Address Problems Involving Complexity
MQALA, which now includes ACPs, can help address complexity by
simultaneously measuring how many variables interact for objects that are
individuals. The variables can be internal or external to the object. The
interactions
can involve variables both within and across levels in measurement
hierarchies.
The variables can act in different combinations. Any interactions can be
positive or
2o negative.
ACPs are a new way to image complexity as it becomes evident in how
individuals
function, respond, and act as agents. Images of complexity based on ACPs can
help users visualize complexity. Visualizing complexity can help make it
understandable.
Images of functional and response complexity should be distinguished fram
images of structural complexity. Brain scans obtained by Computerized Axial
Tomography illustrate structural complexity. In contrast, an ACP ofan
individual's
brain could show how every region ofthe brain interacts with every other
region of
the brain. Such an ACP, which would illustrate functional and response
complexity, would be easier to understand if it were obtained under a given
set of
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
test conditions. ACPs that image functional complexity of individual brains
can be
computed from a series of functional brain images (Section 4.2.7.2).
ACP images of functional and response complexity can be very extensive. ACPs
can have a virtually unlimited number of rows and columns for each dimension.
For example, an image showing functional interconnectivity of brain regions
could
have one row and column for each of the corresponding pixels in the series of
functional magnetic resonance images from which it is computed. Additional
levels could be added for Boolean events.
The ways in which MQALA addresses functional complexity can be viewed from
other perspectives.
One reason why MQALA, including ACPs, is a significant advance in human
history is that although human judgment seems to rely heavily on longitudinal
associations and temporal contingencies, prior art computational methods and
systems for analyzing longitudinal associations have limited functionality. In
contrast, computational methods and systems for analyzing cross-sectional
associations are well developed.
The importance of longitudinal associations and temporal contingencies in
human
' judgment can be illustrated in the context of clinicians judging the effects
of drugs
on patient health. Clinicians often judge how individual patients respond to
drug
challenge, de-challenge, re-challenge, and other changes of dose. Clinicians
often
plan continued treatment of individual patients in accord with such judgments.
Learning from such judgments can be contrasted with learning from conventional
group clinical trials. Sections 2.8.2 and 4.2.2.2 of the parent patent'
application
describe many advantages of using MQALA, a computational method and system
that can supplement human judgment, to help individualize patient care.
Longitudinal associations and temporal contingencies also appear to play
important
roles in the workings of nature. The capacity of brains to learn appears to
have
evolved in a way that allows animals (including people) to learn from temporal
2G
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
contingencies involving stimuli, responses, and rein~orcers. Only recently
have
humans begun to learn by applying the statistical method. Much of human
associative learning also appears to involve temporal contingencies and an
extension of learning capabilities From emotional and motor responses to more
abstract and conceptual entities. Sections 4.2.5 and 4.2.6 of the parent
patent
application includes a discussion o~ how classical conditioning, instrumental
conditioning, and paired associate learning can be analyzed with MQALA and of
how this knowledge can be used to create machines and artiF~zcial systems that
learn.
to
MQALA is an important advance in scientific methodology and for the
application
of technology to achieve human objectives because ACPs can help users
visualize
functional and response complexity; facilitate the creation o~ mathematical
models
of how complex systems Function, respond, and act; as well as the creation o~
15 artificial systems that learn.
2.4.3. ACPs Help Address Problems Involving Nonlinearity
The computational procedures upon which ACPs are based address nonlinearity in
20 at least two primary ways. First, MQALA converts dimensional series into
sets of
dichotomous series using integrated scales as described in Section 4.1.2 and
illustrated in Tables 6 and 7 0~ the parent patent application. The values o~
measures portrayed in ACPs are computed From cross-classifications o~
independent and dependent events as defined on such dichotamous series as
25 illustrated for longitudinal association scores in Section 4.1.1 of the
parent
application. As such, the computational procedures do not assume that e~~ects
o~
independent variables or events on dependent variables or events are
proportional
to independent variable levels.
~0 Second, MQALA can use Boolean independent events and Boolean dependent
events to help determine i~ particular combinations o~ events are associated
more
with other events than with the same variables considered individually. This
use of
Boolean events, described in Section 4.1.11 and illustrated in Table 17 of the
27
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
parent application, addresses the problem of non-additivity as described in
Section
1.2.2.3 of this application.
2.4.4. AGPs Help Address Problems Involving Comprehensive Investigations
Section 1.2.1.2.1.2 and its subsections in the parent application describe a
number
of problems involving comprehensiveness and detail in the context of
evaluating
treatments for health disorders. Section 2.7.1.2.1.2 and its subsections in
the parent
application describe how MQALA helps address these problems. Furthermore,
1o Section 4.2.1.1 of the parent application includes information about how
MQALA
addresses problems involving the emergence of system properties and unique
entities -- problems that are included in Section 1.2.1 also of the parent
application.
ACPs further address the problems involving comprehensiveness as described in
Section 1.2.2.4 of this application because of the capability of ACPs to
provide
quantitative displays of large numbers of interactions simultaneously.
Measurement of the interactions for an individual effectively converts the
interactions into a multidimensional object that can be visualized, graphed,
and
subjected to established quantitative methods such as those ofmorphometrics.
MQALA, including ACPs, is an advance in data processing that helps make
systems biology and discovery science possible. Furthermore, MQALA can be
applied to many types of complex system in addition to biological systems.
This
flexibility of application derives from the fact that MQALA can be applied to
data
far various types of entity much as the statistical method can be applied to
data for
various types of entity. However and again, a fundamental distinction between
the
two methods is that MQALA is specifically designed for application to
individual
entities while the statistical method is best suited for collective entities.
3p 2.4.5. ACPs Help Address Problems Involving Detailed Investigations
There are at least two ways that ACPs and MQALA can help address the need for
detailed investigatians. First, ACPs can address interactions between and
among a
?~
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
virfiually unlimited number of variables. This can, for example, help users
avoid
problems fihat can result when investigators lump many health variables into a
single health measure.
Second, each score in an AGP typically would be a score that summarizes a
multidimensional array of scores. The dimensions of arrays correspond to
analysis
parameters such as IV level, DV level, episode length, episode criterion, and
persistence. users of AGPs and MQALA would be able to drill down into these
arrays to examine in great detail the interactions shown in ACPs.
2.4.6. Addressing the Need fio Investigate AlI Five Types ofProblem as a Set
MQALA is an integrated set of data processing fools built around core
computational measurement methods and providing users with many options about
how to proceed. These core computational methods measure longitudinal
associations between and among variables and events for individual entities
that
are complex systems. As such, MQALA addresses all of the five problem areas -
individuality, complexity, nonlinearity, comprehensiveness, and defiail - as a
set.
Tn practice, all things cannot be investigated at once. Investigators are
limited by
the number of variables that can be measured simultaneously, by the temporal
resolution of data, and by the number of repeated measures that can be
included in
particular investigations. Computer resources for analyzing data are limited.
This is
one reason why users of MQALA and AGPs will be limited in the number of
2S analysis parameters, analysis paramefer levels, and Boolean events that can
be
included in particular investigations.
Despite such limitations, MQA>~A facilitates investigations that are both more
comprehensive and detailed than investigations conducted with conventional
data
3o processing procedures. As such, MQALA can help make better use of data that
can
be collected now. In addition, several uses of AGPs are designed to help
address
the research strategy dynamic involving detailed and comprehensive
investigations. As examples, Section 4.8.8 describes how ACPs can be used to
?9
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
measure interactions involving different types of action. Section 4.8.11
describes
use of ACPs to investigate nested systems. Section 4.8.12 describes use of
AGPs to
distinguish episodes of action. Section 4.8.4 describes how AGPs can be used
to
fingerprint individuals, fingerprints that can be used to develop
classifications of
individuals into groups that axe more homogeneous with respect to how they
function and interact with their environments. Such uses provide tools both
for
partitioning the subject matter of science and for examining how the parts
interact
to form complex systems.
1V.LQALA is based on a tenet that is central to science: If you want to
investigate
something scientifically, measure it! ~IQALA measures the interactions between
and among variables and events that help reveal how complex individual systems
of many types function, regulate, and sustain themselves as well as how they
respond to and act upon their environments.
The above objects and other objects, features, and advantages of the present
invention are readily apparent from the following detailed description of the
best
mode for carrying out the invention when taken in connection with the
accompanying drawings.
BRIEF DESORIPTION OF THE SEVERAL VIEWS OF THE DRAWING
Figure 1 illustrates steps to construct AGPs.
Figure 2 portrays average strength of a longitudinal associafiion measure as a
function of delay for interactions involving GnRH and P-LH.
Figure 3 portrays average strength of a longitudinal association measure as a
function ofdelay for interactions involving GnRH and J-LH.
Figures 2 and 3 are described in Section 4.9.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
DETAILED DESCRIPTION OF THE INVENTION
This invention is a platform data processing technology that applies to
various
types ofobjects and actions, has various computational features, and many
uses.
This section provides a detailed description of ACPs. The first major
subsection
illustrates types of objects that can be represented by ACPs. Objects are
distinguished from their environments.
The next major subsection illustrates types of action that can be manifested
by and
upon various objects of investigation. Several measurement technologies yield
data
that are particularly well suited for processing to yield ACPs. Major
components of
established sciences and disciplines can be investigated with ACPs.
Subsequent major subsections identify various features that can be used to
compute
ACPs and types of scores and measures that can be portrayed by ACPs. These
features, scores, and measures will be presented primarily by reference to
specific
sections of the parent patent application. ACPs themselves can be portrayed in
various ways that will be illustrated in a major subsection.
Additional major subsections illustrate uses of AGPs, cover databases that
include
ACPs, and statistical analyses of two or more AGPs. The final major subsection
presents several examples of ACPs.
4.1. Objects ofInvestigation and Their Environments
ACPs can be used to investigate objects that are complex systems of many
types.
Systems are considered to be complex when they have many parts, variables, or
manifestations that can interact. A system also is considered to be complex
when
many things in its environment can act upon it. Here are some examples of
complex systems that can be investigated with ACPs. These examples are not
exhaustive or mutually exclusive.
~1
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4.1.1. Organisms Including Persons
An organism has been defined as a complex structure of interdependent and
subordinate elements whose relations and properties are largely determined by
their function in the whole. Organisms carry on the activities of life.
Organisms are
considered to include persons.
4.1.2. Portions of Organisms
Portions of organisms include cells, tissues, organs, systems, biochemical
pathways, and biopathways. Such portions of organisms have differentiated
structures or functions.
4.1.3. Economies and Investment Markets
Economies and investment markets are complex systems that can be objects of
investigation with MQALA including ACPs. Such investigations are particularly
feasible because there are vast amounts of readily available time series data
for
these objects.
Scientific investigations of how economies and investment markets function can
be
particularly valuable because of the way improvements in economic policy can
affect society and the way accurate predictions can affect investment profits.
4.1.4. Machines, Processes, and Other Man Made Systems
People make machines, processes, and other complex systems that can be objects
of ACPs. This can be illustrated with refinery processes that can be monitored
in
terms of inputs and outputs for product streams. Such monitoring can produce
time
series dafa that can be analyzed with MQALA. Resulting measures of interaction
can be used to identify how inputs can be controlled to help optimize reFmery
processes.
32
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Other examples of man made systems include healthcare, banking, and soFtware
systems whose behavior can be monitored with time series data.
S 4.1.5. Systems Consisting of Two or More Individuals
ACPs can be used to investigate reciprocal interactions of objects involving
two or
more individuals that may interact to form compound objects. This can be
illustrated with episodes of dance in which the object is a pair of dancers
and the
actions are the movements of both dancers (Section x.2.1). AGPs can be used to
describe the nature and extent to which movements of dancers are coordinated.
Individuals farming compound objects need not be of the same type. For
example,
a horse and a rider or a man and a machine form compound objects.
IS
The same approach could be extended to compound objects with more than two
members such as teams. Numbers of variables and numbers of Boolean events that
could interact would tend to increase rapidly with the number of individuals
in
compound objects.
x..1.6. Papulations Investigated as Wholes
Section 2.6 of the parent application describes how MQALA can be applied to
objects that are populations investigated as wholes. Investigating populations
as
wholes is distinct From malting inferences about populations from samples of
populations' individuals.
The section just referenced illustrates the research strategy of investigating
populations as wholes in the context of epidemiology. This strategy can, for
3p example, use environmental variables such as measures of air pollution that
are
considered to affect whole populations togefller with population variables
such as
rates of death or hospital admission. ACPs For populations considered as
wholes
33
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
could be based for many variables considered to affect populations as wholes
and
various population rates and proportions.
x.1.7. Nested Systems
Complex systems often can be investigated as series o~ individuals with
differing
degrees of inclusiveness. Inclusiveness involves physical, functional, and
conceptual boundaries identified or set by investigators to help distinguish
that
which is within systems from that which is part of systems' environments.
Thus,
for example, investigators of biological systems often distinguish endogenous
from
exogenous substances.
Here is an example of a nested system in the context of investigations of
living
things in which each successive individual can be considered to be nested
within
the next level: cell, organ, organism, and social system. Another example in
the
context of biological systems is neuron, brain, and organism.
Here is an example of a different type of problem that can be addressed using
nested systems: attempting to predict the price of an individual company's
stock.
2o Starting from the least inclusive level of investigation for this
particular example,
the company's stock price can be investigated in terms of its own periodic
measures of business performance. Then the stock's price could be investigated
in
terms of various stock sectors including the sector of which the stock of
interest is
a member. Then the price of a broad range of stocks across many sectors could
be
investigated in terms of other potential investments in a particular country's
economy such as bonds, real estate, commodities, and collectibles. Finally the
performance of a particular country could be investigated in the context of
periodic
measures of how other countries perform.
Activity at each level of a nested system can be expected to have effects on
the
activity of included component individuals. Thus, for example, the performance
of
the economy of the United States in the world economy could be expected to
have
effects on stock prices for individual companies.
3~
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Similarly, individual companies' perFonnances could be expected to have
efFects
as agents on more inclusive levels in nested systems. Thus, for example, the
performance of a major United States company could be expected to have some
impact on the performance of the economy of the United States in the
environment
of the world economy.
AGPs can be used to investigate reciprocal interactions between individuals
and
their environments when the systems are defined to have the different degrees
of
inclusiveness that are characteristic of nested systems. This can be done when
measures of action for one component of a nested system become variables for
computing ACPs for other more or less inclusive components of the nested
system.
Sections 4.1.17 and 4.4.3.8 of the parent application describe predictive
indices,
which are another component of MQALA. Predictive indices can be an efficient
and operational way to summarize the effects of many independent variables or
predictors so that many types of action can be summarized in one variable that
can
in turn be used for computing ACPs for nested systems.
Here is an example of how predictive indices could be used to investigate a
system
consisting of a neuron nested in a relatively simple brain. Suppose that it is
feasible
to obtain time series data on variables that may influence the rate at which a
neuron
ores together with data for the neuron's firing rate. The predictive index
feature of
MQALA could be used to summarize the effects of two or more variables that
2~ affect firing rate. Then the predictive index could be included as a
variable in an
ACP that also includes variables describing action of the brain of which the
neuron
is a part. Such strategies could be used to help elucidate reciprocal
interactions
involving nested systems.
35
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4.1.8. Additional Types of Systems
ACPs can be computed for systems that can be investigated with time series
data
for variables that fluctuate in level or recur over time. Examples include
social
systems, ecosystems, and weather systems.
~.2. Aet10115
Various types of objects that can be represented by ACPs manifest actions of
many
types. The following subsections illustrate various types of action that can
be
measured to yield data that are processed to construct ACPs. Resulting ACPs go
beyond levels of variables from which they are computed to reveal how
variables
interact to form functioning systems. Procedures for obtaining data that are
to be
processed to construct ACPs can be specified in protocols.
Measurement technologies, here considered as they measure various types of
action, axe improving rapidly, often in several ways simultaneously. Here are
examples of different ways in which measurement technologies are improving.
New measurement technologies are measuring things that were never measured
before. For example, modern imaging technologies can measure brain activity
such
as areas of the brain that are activated while performing mental arithmetic.
Rating
scales are being developed to operationalize concepts at high levels of
measurement hierarchies, concepts such as health-related quality of life.
Measurement technologies are becoming more sensitive. For example, they can
detect and measure concentrations of parts per billion or less.
Some measurement technologies can measure thousands of variables
simultaneously. Microarrays that are used in biotechnology can measure
thousands
of variables on one chip.
3G
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Some measurement technologies are achieving high temporal resolution. Temporal
resolution refers to the number of repeated measurements per unit time. Some
processes can be measured repeatedly to yield time series data with many
measurements per' Second.
S
Measurement devices are being miniaturized. MicroElectroMechanical Systems
(MEMS} for sensing flow, pressure, and chemicals are being developed that can
operate from within the human body.
1D Measurement devices are being Internet enabled. For example, Internet
enabled
medical monitoring devices can report health variables from almost anywhere.
Most improvements in measurement technologies, inventive as they are, are
conventional in an important way. That is, most conventional measurement
15 technologies measure variables one by one as distinct from how the
variables
interact to form complex functioning systems. In contrast, this invention
measures
how variables interact to form complex functioning systems.
This invention is a computational method and system. The input for this
invention
20 can comprise much output from conventional measurement technologies. The
output from this invention comprises quantitative descriptions of how complex
individual systems function. As such, this invention increases the value of
many
conventional measurement technologies.
25 Here are some examples of types of action that can be measured and serve as
input
for this invention. The following types of action are not mutually exclusive
or
exhaustive. The Following subsections include information about how ACPs for
various types of action can be useful.
30 4.2.1. Movement
ACPs were inspired by the task of analyzing data collected by motion capture
technology. One such technology that can be applied to humans uses 36
reflective
~7
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
markers on various parts of the body together with Eve video cameras to
capture
three dimensional motion data. Data can be captured during movement protocols
that capture episodes of movement such as walking or a short series of
repeated
golf swings. Software that is part of the motion capture technology computes
values of many variables comprising measures of linear and angular velocity
and
acceleration. Motion capture technology can result in extensive databases of
high
temporal resolution data with thousands of variables. Such databases are
descriptive of particular episodes of movement for particular individuals.
to ACPs can be computed from such databases. Such ACPs can show how every
variable interacts with every other variable. ACPs for episodes of different
behavior and for different individuals would be expected to have
characteristics
that could be assessed with morphometric analyses to identify salient features
that
distinguish various ACPs.
ACPs for movement will be used to illustrate coordination. Suppose that a
particular movement protocol calls for a series of three repeated golf swings.
It is
anticipated that an expert golfer would have a well-defined ACP with many
large
magnitude scores. Such scores would indicate that particular movements are
highly
coordinated with other movements. Scores at specific values of delay would
indicate precise sequences of movement. In contrast, the ACP of a novice
golfer
under the same conditions would be anticipated to be relatively flat. Movement
sequences would not be as specific and repeatable.
ACPs also could be used to picture the effects of various interventions.
Consider
again the ACP of an expert golfer. Intoxication with a drug that alters
behavior
could be expected to flatten such a profile.
ACPs can be used to investigate coordination for actions in addition to
movement.
38
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4.2.2. Physical and Electromagnetic Action
Physical and electromagnetic action involves properties that are part of the
subject
matter of physics. Included would be measures o~ pressure, flow, wavelength,
intensity, and voltage. Some of the most interesting properties of complex
systems
appear to emerge from how basic physical actions are organized and interact.
Biological activity often is monitored with technologies such as
electroencephalography and electrocardiography. Such procedures can yield many
variables in the frequency and spatial domains. Such variables can be analyzed
with ACPs.
4.2.3. Chemical or Biochemical Action
Changing levels or concentrations o~ chemicals can be measured repeatedly and
may indicate chemical reactions or interactions. ACPs can be used to help
investigate how such actions may be coordinated.
Biochemical reactions underlie much of the life sciences and medicine.
Biologically active substances include endogenous products of gene expression
such as hormones, neurotransmitters, receptors, and messenger molecules, as
well
as exogenous substances such as drugs and other chemical exposures. Many of
the
laboratory measures used in medicine re lest biochemical action.
4.2.4. Biological Action
Biological action becomes evident in living things. Although biological action
includes physical and biochemical manifestations of living things, a new, more
inclusive view of biological action is emerging. This is illustrated by the
web site
for the Institufe of Systems Biology, "systems biology studies the complex
interaction of all levels al' biological information: DNA, mRNA, proteins,
~9
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
functional proteins, informational pathways and informational networks to
understand how they work together."
The systems approach to biology can be contrasted with the more conventional
approach that tends to investigate components of biological systems one by
one.
The systems approach explicitly recognizes that some of the most interesting
and
important properties of biological systems, such as immunity, are emergent or
systems properties.
1o The systems approach to biology also can be applied to portions of
biological
systems. For example, attention, memory and learning can be considered as
emergent or systems properties for brains or nervous systems.
MQALA can be applied most directly to physiological and environmental
1 S variables that can fluctuate in level over time and events that can recur
for
individuals. Although MQALA is not well suited to investigate structures
themselves, MQALA can be applied to investigate how structures function.
4.2.5. Emotional, Mental, and Behavioral Action
Although many types of systems can be said to behave, this section focuses on
psychology. The science of psychology investigates the behavior of animals and
people, primarily for individuals as contrasted with groups. Psychology is
considered to include investigations of emotional and mental action. As with
biology, the subject matter of psychology usually has not been investigated as
systems. MQALA, including AGPs, constitute a new set of analytic tools for
investigating behavior and learning phenomena from a systems perspective.
Much of the subject matter in the psychology of learning can be considered to
involve longitudinal associations (temporal contingencies) involving stimuli,
responses, and reinforcers. Section 4.2.6 of the parent patent application
includes
major elements of an analysis of classical conditioning, operant conditioning,
paired-associate learning, and associative learning in terms of changes in one
type
~0
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
o~ longitudinal association in the presence of another type of longitudinal
association. Furthermore, Section 4.2.6.4 0~ the parent application
illustrates how
such analyses can be applied to create artificial systems that learn.
Artificial
learning systems or robots can be considered as embodied models o~ learning
systems.
Mental disorders are considered to involve emotional, mental, and behavioral
action.
4.2.6. Social and Economic Action
Measures o~ social and economic action are a foundation for scientific
investigations of collective entities such as societies and economies.
Economic
measures often are applied repeatedly to yield time series data that can be
processed to construct AGPs.
4.2.7. Action Measurement Technologies
Major advances in sciences and disciplines such as medicine that apply science
often depend on advances in measurement technologies. Some of these
technologies are being singled out for speoial attention.
New measurement technologies tend to yield more data than people have been
able
to understand. This invention can help process much of this data to facilitate
understanding o~ how complex systems function and interact. As such, this
invention can increase the value of these measurement technologies by serving
human needs.
4.2.7.1. Biochemical Measurement Technologies
Flew developments in biochemical measurement technology include microarrays
or chips that can measure thousands of products of gene expression at one
time.
Any of these gene products may interact with any other gene product-
interactions
41
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
that may indicate up or down regulation of biochemical pathways or
biopathways.
Huge numbers of interactions can be investigated simultaneously when such
chips
and related procedures can be applied repeatedly and the resulting data are
used to
construct ACPs.
ACPs can be used to help quantify, discover, and describe protein-protein
interactions. Quantification o~ protein-protein interactions would help make
life
sciences more productive. Discoveries and detailed descriptions made with ACPs
could be objects of patent applications that identify biochemical pathways
that may
be new targets for drug development.
Furthermore, ACPs can be used to analyze protein-protein interactions. Section
x..8.9.2 describes how ACPs can be used alone and together with experimental
procedures to help investigate the temporal criterion of causal and other
predictive
associations. Interactions can be analyzed as functions of analysis parameters
used
to construct AGPs. Interactions can be analyzed when operators such as AND,
OR,
NOR, XOR and NOT are applied to sets of independent and dependent events to
define Boolean events. Such analyses can contribute to a detailed
understanding of
how organisms function.
Exogenous substances can be used as experimental probes to investigate
biological
systems. Here are two distinct and novel research strategies that involve the
use of
experimental probes and ACPs.
The first o~these strategies would involve administration of the experimental
probe
repeatedly, possibly at variaus levels or doses, while many biochemicals are
being
measured at the same time. The resulting data would be used to constn~ct ACPs
that include the probe as a variable. This strategy would help reveal how the
probe
may interact with all the biochemicals.
3Q
The second of these two research strategies involves constructing ACPs under
different treatment conditions, preFerably For the same individuals. For
example,
separate ACPs can be constructed for the ''on treatment" and the ''oFf
treatment"
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
conditions. Differences in such ACPs would help reveal how treatment may
affect
any of many interactions portrayed in such ACPs. Unlike the First strategy,
treatment would not be included as a variable in ACPs constructed with the
second
strategy.
The two research strategies just described are fundamentally different. The
first
strategy would reveal how treatment may affect levels of all the biochemicals.
For
example, the first strategy could help reveal how treatment may affect insulin
and
glucose levels.
In contrast, the second of these strategies would help reveal how treatment
may
affect any of the interactions between or among the biochemicals. For example,
treatment may help restore a disordered interaction between insulin and
glucose
levels. Both research strategies could facilitate drug development by
quantifying
I S interactions that may indicate efficacy or possible safety concerns.
4.2.7.2. Functional Imaging
Functional magnetic resonance imaging and Positron Emission Tomography can
measure biological activity in, for example, brains. Section 4.2.4 of the
parent
application explains how data from a series of such aligned images can be used
to
form serial pixel or serial region of interest variables. One such variable
would
correspond to each pixel or region of interest. Regions of interest could be
delimited on the basis or shared structural or functional characteristics.
Corresponding computational procedures could be applied to form serial voxel
(volume element) variables.
ACPs can be computed using serial pixel, serial voxel, or serial region of
interest
variables and presented graphically as images. Such images would show how
activity in each pixel, voxel, or region of interest interacts with activity
in every
other pixel or region of interest that is included in the ACP. Section 4.5
includes
more information about portrayal of images and displays based on AGPs
constructed from functional imaging data.
43
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Functional connectivity within brains that is revealed by ACPs and derivative
images could be investigated experimentally and non-invasively with
transcranial
magnetic stimulation of particular brain regions.
4.2.7.3. MicroElectroMechanical Systems (MEMS)
MEMS are super-miniaturized machines that can perform combinations of sensing,
processing, communicating, and acting upon information. Fox example, so-called
lab-on-a-chip MEMS are being developed that can operate from within human
bodies.
MEMS will vastly increase amounts of time series data that can be analyzed
with
ACPs. This includes analyses that can be implemented within MEMS themselves.
1S
4.2.7.4. Instrumentation far Psychophysics and Psychometrics
Psychometrics involves mental measurement. Psychophysics is concerned about
the effects of physical processes such as wavelength and intensity on mental
2o processes of organisms. Measurement technologies in these areas, many of
which
are computerized, help make the subject matter of psychology amenable to
scientific investigations.
4.2.7.5. Performance Measures
Performance is measured in terms of things such as speed, accuracy, and skill.
Computerized performance measurement technologies often are particularly well
suited for yielding data that can be analyzed with ACPs.
3p 4.2.7.4. Rating Scales and Surveys
Rating scales and surveys that often are based on selF report are benefiting
from
new measurement technologies that improve reliability, validity, and ease of
use
~4
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
for many measures. For example, measures of health related quality of life are
being developed that help provide a common metric for evaluating the effects
of
many treatment across many disorders.
Many rating scales and surveys can be applied repeatedly over the Internet.
This
facilitates the collection of repeated measures data that can be processed to
construct AGPs.
x.2.8. Sciences and Disciplines
Major sciences and disciplines such as chemistry, biology, psychology,
sociology,
economics, and medicine can be considered to focus on certain combinations or
clusters of objects and actions. As examples, psychology tends to focus on
behavior of individual organisms. Medicine tends to focus on biochemical and
behavioral signs and symptoms of individual organisms with health disorders.
The names of various sciences and disciplines are used to identify areas of
application of MQALA including AGPs.
4.3. Computational Features
Various computational features ofMQALA can be used to define independent and
dependent events. These events are determined to be present or absent over
different times thus forming dichotomous series or sets of dichotomaus series
used
to compute AGPs. Table 1 lists section numbers of the parent application that
describe these selected features. Section ~.~ of the parent application
illustrates
most of these Features with an example.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Table 1. Computational features by section number in parent application.
i FEATURE ~ SECTION NUMBER OF PARENT
APPLICATION
Variable level 4.1.2
Episode Length 4.1.7
Episode Criterion 4.1.8
I
Delay 4.1.9
Persistence 4.1.10
Boolean Events 4.1.11
Delay and Persistence After
Boolean ' 4.1.12
i Events
Transition Events 4.1.13 '
Other Procedures Applied to
Farm 4.1.14
i Additional Dichotomous Series
Variable level is a required analysis parameter when data for a variable are a
dimensional series and users ofMQALA select to investigate more than two
levels
of the variable. Episode length and episode criterion are optional analysis
parameters that can be applied to define independent andlor dependent events.
Delay and persistence are optional analysis parameters that can be applied to
define independent events.
Boolean events and transition events can be applied to define independent
events,
dependent events, or both. Boolean events and transition events add columns
and
rows to dimensions ofACPs.
4.4. Measures Portrayed in ACPs
The parent application describes various measures that can be portrayed in
ACPs.
Parent application Section 4.1.1 describes how longitudinal association scores
are
computed. Parent application Section 4.1.5 describes how longitudinal
association
scores can be converted to bene6t/harm scores ~or evaluation research. Parent
application Section 4.1.6 describes computation oC fihree related measures of
strength of longitudinal association. Parent application Section 4.1.3
describes how
longitudinal association scores and beneFifi/harm scores can be summarized.
~tG
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Analogous procedures can be used to summarize strength of longitudinal
association measures.
4.5. Portrayal of ACPs
ACPs or portions of ACPs can be portrayed with various types of tables,
figures,
graphs, images, displays, or interactive displays. Section ~.9 of this
application
gives several examples of ACPs and portions of ACPs that are portrayed as
tables.
Such tables also can be portrayed graphically.
Portrayal of ACPs as images and displays will be illustrated in the context of
functional imaging as described in Section x..2.7.2. Such images and displays
could
be used to investigate functional connectivity of brain regions in ways that
are
useful for research and diagnosis. These images and displays could be color
coded
so that one range of colors represents degrees of excitatory activity
(positive
scores) and another range of colors represents degrees of inhibitory activity
(negative scores).
ACPs derived from series of Functional images can portray vast amounts of
information. This is true, for example, when ACPs are computed from pixel or
voxel data with one variable for each pixel ax vaxel. Such ACPs could have one
row or column for each pixel or voxel functioning as an independent variable
and
one row or column for each pixel or voxel functioning as a dependent variable.
However, such AGPs would not portray information in accordance with familiar
anatomical shapes.
Functional cannectivity of brain regions could be portrayed in accordance with
familiar anatomical shapes by imaging or displaying ACP information one row or
column at a time. This would yield a type of derivative image in which a
particular
pixel, voxel, or region of interest could be blank while measure values for
all other
cells in a column or row could be shown with color coding in there
anatomically
correct positions. Resulting images for particular brain regions would image
~7
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
functional connectivity of the particular regions with respect to all other
regions
included in the images.
The type of image just described could be poutrayed in an interactive format.
Suppose one wants to image functional interconnectivity using two-dimensional
cross-sections of brains. Further suppose that one can use a cross-hair type
pointer
on a computerized image to identify a particular brain region. Once
identified, the
image would show how action in the particular region is associated with
activity in
every other region. For this type of interactive display, the required scores
could
either be stored in a potentially very large ACP or the required scores could
be
computed on demand.
~.6. Databases that Tnclude AGPs
1 ~ ACPs or portions of ACPs can be stored in databases, with or without
additional
data. Section ~.7 describes some ways in which these databases can be used.
4.7. Analyses of ACPs with Statistics or Other Quantitative Methods
2o ACPs are sets of measurements that can serve as input or be operated upon
by
statistics and other quantitative methods. Quantitative methods can be applied
to
individual ACPs, two or more ACPs for an individual, AGPs for two or more
individuals, or portions oFACPs.
2S ACPs can be said to give shape to interactians. Shapes can be measured with
morphometry.
ACPs can be analyzed statistically. For example, twa or more ACPs with
corresponding structures can be averaged. ACPs From two or more individuals
can
30 be used to describe groups or to make inferences about populations.
Science generally Favors parsimonious explanations. Various quantitative
methods
such as factar analysis, discriminative analysis, and cluster analysis aan be
applied
~1S
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
to ACPs. Such techniques, for example, can be applied to help reduce the
number
of variables that need to be measured while still capturing interactions that
are of
interest.
4.8. Uses of ACPs
MQALA, including ACPs, have many uses in addition to applications to various
objects, actions, subject matters, and combinations thereof. The following
subsections identify and describe same such uses. Together, such applications
and
uses form an extensive and varied tool kit for addressing problems involving
repeated measures and time series data. Many problems are best addressed with
time series data alone or in combination with cross-sectional data.
4.8. ~ . Use of ACPs to Measure Tntemal Control
ACPs that are constructed with variables that include two ox more variables
considered to be internal to the object represented by the ACP can be said to
include measures of internal control. Terms used to describe internal control
vary
by scientific discipline. For example, some investigators speak of regulatory
control in biological systems. In this context, the variables may measure
endogenous substances such as hormones, neurotransmitters, messenger
molecules, or other components of biochemical or signaling pathways. Internal
control includes biochemical pathways or biopathways. Psychologists may use
terms such as self control.
Systems that include feedback or feed forward mechanisms often can be said to
exhibit internal control.
4.8.2. Use ofACPs to Measure Responses to Environments
ACPs that are constructed with variables that include at least one variable in
the
environment of the object that is being represented by fine ACP can be said to
include measures of response to the environment. More specifically, those
portions
~9
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
of such ACPs where environmental variables) function as independent variables
partray responses to environments.
Treatments are an important subclass of environmental variables. Portions of
ACPs
that are responses to treatments generally would be portrayed with
benefitlhann
SGOreS.
x.8.3. Use of ACPs to Measure Actions on Environments
Portions of ACPs correspanding to environmental variables Functioning as
dependent variables can be said to measure actions of objects represented by
ACPs
on their environments. In such cases, objects can be said to be functioning as
agents acting on their environments.
X1.8.4. Use of ACPs to Fingerprint Tndividuals
The section of this application that is titled "Brief Summary of the
Invention"
describes how ACPs can be used to fingerprint individual objects in terms of
how
they function or interact with their environments.
x..8.5. Use of ACPs to Portray Test Results
ACPs that portray test results can be descriptions of how their objects
function and
interact with their environments. Test results can be presented for many types
of
?5 independent and dependent events simultaneously. Abnormal results on such
tests
could be diagnostic of disorders such as health disorders.
The tests from which ACP test results are obtained generally would be
conducted
under standardized conditions. Test conditions could be specified in test
protocols.
Test protocols would need to speciFy things such as type of object that is
tested,
variables used to construct ACPs and how they are to be measured, how any
environmental variables would be controlled, all scoring Features used to
transCoan
data and define events, as well as the type of'measures portrayed in ACPs.
SO
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
ACPs that result from tests could be accumulated in databases that include
ACPs
from many individuals. These databases could be analyzed to help distinguish
normal from disordered test results and to classify individuals with distinct
or
S disordered test results.
x.8.6. Use ofACPs to Identify Predictors ofDisorder
Section X4.8.5 describes how ACPs can be used to portray test results that can
in
1o turn be analyzed to develop classifications of abnormal test results that
indicate
disordered functioning. Such classifications can be used to help identify
predictors
of disorder. For example, this approach could be used to search for single
nucleotide polymorphisms that are predictive of abnormalities in the
regulation of
glucose metabolism.
4.8.7. Use of ACPs to Identify Predictors of Differential Response
AGPs that are based on multiple treatment and response variables can provide
detailed descriptions of responses to treatments. Such ACPs for groups of
patients
2o can be used both to help classify responses to treatment and to help
identify
predictors of differential response. For example, this approach could be used
to
search for single nucleotide polymorphisrns that are predictive of
differential
responses to drugs.
~.8.$. Use of AGPs to Measure Interactions Involving Different Types of Action
Section X1.2 of this application illustrates several broad categories of
action that can
be investigated with ACPs. ACPs that are based on variables that measure
different
types of action can be used to investigate interactions involving actions of
different
3Q types.
A number of scientific disciplines focus on interfaces between various types
of
action. Biochemistry is at the interface of chemistry and biology.
Psychophysics is
SI
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
at the interface between physics and the psychology of perception. Social
psychology is at the interface between behavioral and social action.
Section 1.2.2.2 illustrated hierarchies of measurement in the context of
complexity.
ACPs can be used to investigate interactions involving measures at different
levels.
For example, ACPs could be used to investigate interactions involving
laboratory
measures used in medicine and self reported symptoms.
4.8.9. Use of ALPS to Help Distinguish Causal from Non-causal Associations
ACPs are quantitative descriptions of associations. ACPS generally portray how
each type of event, defined on variables, is associated with every other type
of
event defined on the variables. Associations often do not indicate causal
relationships.
is
Science often quests for causal relationships. MQALA and ACPs can be used to
explore for causal relationships. In addition, MQALA and ACPs can be used
together with the experimental method to help confirm causal relationships.
Both
strategies will be discussed in turn.
4.8.9.1. Exploring for Causal Relationships
MQALA and ACPs can use the analysis parameter called delay to help evaluate
the temporal criterion of causal and other predictive relationships. Delay is
defined
on variables functioning as independent variables. It is more feasible to
investigate
delay with time series data in which measurements are obtained periodically
after
equal units of time.
The default value of delay is 0. When delay is set equal to 0, the
associations that
3p are measured are among events present at the same time. When delay is set
equal
to 1, associations involve dependent events occurring one time unit after
independent events. Users generally would be able to select additional integer
values of delay.
52
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Typically the upper and lower portions of ACPs, as indicated by the symbols
"o"
and "~" respectively in Table 1, will not be symmetric, especially with
nonzero
values of delay. These asymmetries can be used to help investigate the
temporal
criterion of causal arid other predictive associations. According to the
temporal
criterion, causes or predictors should come before effects or predicted
events.
In the absence of any associations, all scores above or below a concordant
diagonal
in ACPs would be expected to hover around zero. This could be illustrated by
computing ACPs from data consisting of random numbers or random normal
deviates.
An asymmetry can be illustrated simply with variables A and B. Suppose that
the
value of a strength of longitudinal association measure is substantially
larger when
Variable A functions as an independent variable and Variable B functions as a
dependent variable than vice versa. This would indicate that Variable A is a
stronger predictor of Variable B than Variable B is a predictor of Variable A.
In
general, large asymmetries would provide more revealing information about the
direction of causal and other predictive associations. Asymmetries also can be
examined for Boolean events.
Asymmetries in AGPs can be examined in different ways. Figures 2 and 3,
further
described in Section 4.9, illustrate asymmetries using data from reproductive
endocrinology.
Asymmetry tables or graphs, which are computed from ACPs, can be used to
investigate the temporal criterion. One way to construcfi an asymmetry table
or
graph is to subtract values from one portion of an ACP from corresponding
values
in the other portion of the ACP. Asymmetry with respect to, for example, the
cell
corresponding to IV 6 and DV 3 can be examined by subtracting the value in
this
cell from the value in the cell corresponding to IV 3 and DV 6.
53
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
ACPs and asymmetry tables or graphs are useful tools for exploratory data
analyses that involve causal or other predictive associations. This type of
exploratory data analysis or data mining can be conducted using ACPs that
summarize across a range of nonzero positive values of delay. Such ACPs show
how each variable may be associated to each other variable in a single ACP or
asymmetry table or graph.
In addition, the temporal criterion of causal or other predictive associations
can be
investigated using a series of ACPs and asymmetry tables or graphs that
1D summarize and quantify associations as functions of various values of
delay.
A series of delay specific AGPs and asymmetry tables or graphs can be used to
investigate cascades of events in complex systems. For example one variable
may
be associated with increases or decreases in a second variable after some
delay. In
15 turn increases or decreases in the second variable may be associated with
increases
or decreases in a third variable after some additional delay. In general, more
delayed effects are apt to be weaker because they are mediated by a series of
more
immediate effects that may dampen more delayed effects. It also appears that
ACPs and asymmetry tables can be used to investigate positive and negative
2o feedback loops.
The next section describes how MQALA and ACPs can be applied to experimental
data to help confirm causal relationships. However, for some individuals such
as
large-scale economic or environmental systems, it may not be feasible fo
conduct
25 experiments that require isolation of independent variables. For such
individuals, it
may be more feasible to investigate causality by applying MQALA and ACPs with
wide ranges of scoring options and data that are as comprehensive (measure
many
relevant variables) and detailed (avaid composite variables) as feasible. In
addition, it generally would be advantageous For the repeated measurements to
be
30 collected periodically, frequently, and often over long periods of fime.
S:
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4.8.9.2. Conf rming Causal Relationships
Scientists often use experiments in attempts to conl'inn causal relationships.
Experimental strategies and procedures include randomization; isolation of
independent variables; and blinding of subjects, observers, and data analysts.
MQALA is entirely compatible with such strategies and procedures when they are
applied during data collection. In fact, MQALA often enables new ways of
applying such strategies and procedures by expanding the range of types of
data
that can be analyzed effectively by computational procedures. For example,
MQALA often helps make it feasible to randomize doses of treatment to
different
time periods for individual patients. In addition, two or more dose levels
(which
may or may not include a placebo dose of zero) can be used for each patient.
The conventional experimental strategy that often is applied in conjunction
with
the statistical method is to assign individuals to separate groups such as
treated and
untreated patients. An alternative strategy that is facilitated by MQALA is to
control variable input signals for individuals to examine how a host of other
variables interact with the signal. ACPs for two or more individual subjects
could
then be analyzed statistically.
4.8.I0. Use of ACPs for Data Mining
Section 4.2.1.2 of the parent application describes the application of MQALA
to
data mining. Section 1.1 of this application describes the interactions that
ACPs
quantify as becoming evident as patterns in repeated measures and time series
data.
ACPs can be used to mine for such patterns.
ACPs can be a major tool for discovery science as described in Section
1.2.2.4.
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4.8.11. Use of ACPs for Tested Systems
Section 4.1.7 illustrates nested systems and how they can be investigated with
AGPs. AGPs can be important tools for investigating interactions involving
physical or conceptual entities with different degrees of inclusiveness.
4.8.12. Use of AGPs to Distinguish Episodes of Action
As described in Section 2, scientific investigations need to be limited in
scope. As
illustrated in this application, particular ACPs are limited to particular
types of
action measured by particular variables for particular types of objects.
Another
crucial way to limit scope is to limit AGPs to particular episodes of action.
Walk, cantor, trot, and gallop illustrate episodes of locomotion for horses.
Similarly, episodes can be distinguished for different types of action. MQALA
includes tools for helping to distinguish episodes of action.
Section 4.1.15 of the parent application describes sequential analysis of
longitudinal association scores and strength of longitudinal association
measures.
Sequential analysis also can be called iterative analysis.
Sequential or iterative analysis also can be applied to AGPs. With this
procedure,
AGPs would be constructed sequentially over measurement occasions. Measure
values for corresponding locations in AGPs can be plotted as functions of time
or
measurement occasion. Such graphs could be analyzed by looking for inflections
or sets of inflections to help distinguish different episodes.
4.8.13. Use of ACPs for Model Development and Testing
As described in Section 1.2, scientific knowledge often is represented in the
form
of mathematical models. AGPs can be used to inform the model development
56
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
process. ACPs also can be used to test dynamic models by comparing ACPs
produced by models with ACPs for the objects and actions modeled.
~.8.1~. Use of AGPs to Draw Generalized conclusions and to Make Predictions
This application is a child of the parent application with the title
"computational
Method and System to Perforce Empirical Induction." Empirical induction was
defined as involving drawing generalized conclusions and malting predictions
from
data. As such, drawing generalized conclusions and making predictions are
inventive and important uses of AGPs.
Section 1.2 of the parent application presents four criteria for high quality
generalized conclusions and predictions. AGPs can be used as tools to help
draw
high quality generalized conclusions and to make high quality predictions.
X1.8.15. Use of AGPs to Make Scientific Discoveries
Advances in technology for measuring actions have far outstripped our ability
to
make scientific discoveries based on how actions interact. MQALA, including
ACPs, address this problem by measuring interactions with computational
methods
and systems.
X1.8.16. Use of AGPs to Guide Decision-Making
Section 2.8.2 of the parent application describes how MQALA can be used from a
practical perspective in the context of decision-making. Similar arguments
apply to
this improvement on MQALA, AGPs.
X1.9. Examples of AGPs
Tables 2 through 5 and Figures 2 and 3 are based on data from reproductive
endocrinology. For these tables and Cgures, the objects of investigafion are
ewes,
the actions involve hormones, and the use is to investigate internal control.
57
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
The data for Tables 2 through 5 and Figures 2 and 3 are described and reported
in
two publications: Padmanabhan, V., McFadden, K., Manger, D.T., Karsch, F.J.,
and Midgley, A.R. (1997). Neuroendoarine control of follicle-stimulating
hormone
(FSH) secretion. 1. Direct evidence for separate episodic and basal components
of
FSH secretion. Eiadocy°i~aology 138, x.24-X32 and Midgley, A.R.,
MoFadden, K.,
Ghazzi, M., Karsch, F.J., Brown, M.R., Manger, D.T., and Padmanabhan, V.
(1997). Nonclassical secretory dynamics of LH revealed by hypothalamo-
hypophyseal portal sampling of sheep. Endocrine 6, 133-113. The authors kindly
provided access to data described by these publications.
Tables 2 through 4 are based on data for one ewe. Data for these tables were
obtained by assessing five hormone measures every 5 minutes for about 12 hours
-
1~3 repeated measurements of each measure. The hormone measures are
gonadotropin releasing hormone (GnRH), portal luteinizing hormone (P-LH),
jugular luteinizing hormone, portal follicle stimulating hormone (P-FSH), and
jugular follicle stimulating hormone (P-FSH). Portal measures were obtained
from
blood sampled near the pituitary gland.
Table 2 and Table 4 each portray an AGP. The scoring protocol for Tables 2 and
4
is, in brief, as follows. The set of dichotomous series for each variable was
formed
by first computing the standardized residuals from its linear regression line
on time
or measurement number. Next, intervals of z-scores for the residuals were used
to
form 12 dichotomous series for each variable.
In addition, the scoring protocol for Tables 2 and ~. used optianal values of
additional analysis parameters. Ten combinations of episode length and episode
criterion were applied to variables functioning as independent variables and
to
variables functioning as dependent variables. These 10 combinations resulted
from
applying episode length values 1 through ~ and all values of episode criterion
that
are possible given these values of episode length. Delay values of 0, 1, and 2
and
persistence values of 1 and 2 were applied whenever a variable functioned as
an
independent variable. Table 1 identifies sections of the parent application
that
describe these analysis parameters.
58
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Table 2 portrays summary longitudinal association scores. Table 2 includes
summaries of three delay-specific portions of an AGP plus a summary of the
entire
ACf. The summary of the entire ACP is summarized across the three delay-
speciFc portions of the ACP.
Each score in the delay specific sections of Table 2 is summarized across
28,800
delay specific longitudinal association scores. The number 28,$00 is the
product of
12 levels of the independent variable, 12 levels of the dependent variable, 10
combinations of episode length and criterion for the independent variable, 10
combinations of episode length and criterion for the dependent variable, and 2
levels of persistence. Each summary score in the summary ACP portion of Table
2
is summarized across 86,00 scores (3 times the 28,800 delay-specific scores).
Table 2 is a particular portrayal of the entire ACP constructed from the
hormone
data and with the scoring protocol described earlier in this secfion. The
entire AGP
includes 1,728,000 longitudinal association scores - 86,00 for each of the 20
interactions. Each of 20 interactions is evaluated with respect to 8
dimensions that
correspond to analysis parameters.
The first score in Table 2, the summary score for the interaction between GnRH
functioning as the independent variable and P-LH functioning as the dependent
variable, is 76.728. This score is one score from a distribution of scores
with a
mean of zero and a standard deviation of 1. The distribution consists of all
47
scores that are possible given the marginal frequencies of the 2 x 2 table
from
which the score was computed. The 2 x 2 table resulted from the cross-
classi~cation of a particular member of the set of dichotomous series
representing
the independent variable with a particular member of the set of dichotomous
series
representing the dependent variable.
The distribution of longitudinal association scores that includes 76.728 is
shown in
Table 3. The magnitude of this scare (76.728), shown in bold at the bottom of
Table 3, indicates that there is much evidence for a longitudinal association
between GnRH and P-LH. Furthermore, the association is positive--high levels
of
59
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
GnRH are associated with high levels of P-LH. Section 4.1.1 of the parent
application shows how cells of the 2 x 2 table are labeled and how
longitudinal
association scores are computed.
Each interaction portrayed in Table 2 could be investigated in detail by
examining
all or subsets of the longitudinal association scores that are summarized.
This
includes examination of interactions as functions of any or all of the
analysis
parameter levels used in the analysis. The location of the summary score in
the
array of scores that was summarized identifies levels of all analysis
parameters that
1o yield the most evidence for the longitudinal association.
Table 2. A portrayal of the ACp for the hormone data that uses summary
longitudinal association scores.
Summary
of
the
Delay
---
0 Specific
Portion
of
the
ACP
I Independent ~
Dependent
Variable
'
Variable GnRH P-LH J-LH P-FSH i J-FSH
GnRH ' ?6.728 47.287 62.909 22.472
P-LH 76.165 i 48.329 74.014 ' -15.312
J-LH 36.773 i 41.970 28.605 ~ 19.569
P-FSH ~ 74.014 ' 37.956 16.684
61.014
J-FSH 16.503 -15.533 19.253 11.900
I
I Summary
of
the
Delay
= 1
Specific
Portion
of
the
AGP
GnRH 73.359 52.844 ' 31.574
47.991
P-LH 64.184 53.892 ~ 28.745
47.755
J-LH -31.888 34.509 -27.607 19.569
~ P-FSH50.764 51.684 48.622 '
24.488
J-FSH i -19.741 ' -16.764
-19.005 19.529
I Summary
of
the
Delay
= 2
Specific
portion
of
the
ACP
GnRH - 4~.-14~ 52.337 I 33.838
28.408
~ p-LH ~ 51.378 i 28.8_83
-37.334 28.724 I
J-LH -31.888 ' -28.221 16.506
31.184
P-FSH -34.891 -33.156 44.837 31.971
J-FSI-1' -24.279 -14.492 '
-20.389 -21.441
I Summary
of
the
Entire
ACP
GnRH I 52.841 -Ga.909 33.838
76.72$ - I
I
P-LH ' 53.892 71.014 28.883
76.165
J-LH 36.773 ' ~ 1~.5G9
41.970 28.605
' P-FS1161.014 i 48.622 ' 31.971
74.014
J-FSH ' -24.279 ' -21.441
-20.389 19.529 I
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Table 3. Distribution of longitudinal association scores that includes 76.728.
i Longitudinal
Cell
Frequencies
~
i ~ b c d ~ Association i Probabilify
A ~ Score
0 ~ 46 48 ~ 4$ -20.015 3.22270e-11
j
1 45 47 i 49 -17.525 1.45219e-9 i
2 44 46 50 -15.199 3.07138e-8
3 43 45 51 i -13.039 4.06305e-7
4 42 44 ' S2 -11.045 3.779$1e-G
41 43 ~ 53 -9.216 2.6358$e-5
6 40 42 54 ' -7.553 0.00014342$
7 39 ' 41 55 ~ -6.055 D.000G258G7
i
--
' 3$ ~ 40 SG -4.723 ' 0.00223384
8 i
9 37 39 ' S7 -3.556 0.00661879
36 3$ ~ 58 -2.555 D.O1G4G71
11 35 37 59 -1.719 0.0347103
12 34 3G 60 ' -1.049 0.0624304 i
13 33 35 61 ~ -0.544 ' 0.0963616
14 32 34 ' 62 -0.205 ~ 0.128223
~
31 ' 33 ~ 63 -0.032 ~ 0.147626
16 30 i 32 64 O.O1D 0.147482
17 29 31 G5 0.1 G8 0.128129
i ~ 28 30 ' G6 ' 0.490 0.09695_97
18
i 27 ~ 29 ~ 67 D.979 0.0639797
19
26 28 68 ~ 1.633 0.036$354
21 25 27 69 2.452 0.0185D67
22 24 26 70 3.437 0.00811169
23 23 25 71 4.587 0.00309963
24 22 ~ 24 ' 72 ~ 5.903 ~ O.DD 103142
21 23 ~ 73 7.385 0.000298404
i 20 22 74 9.032 7.49112e-5
26
27 19 i 21 75 10.$44 1.62770e-5
28 18 20 ~ 7G j 12.822 3.05194e-6
29 ' 17 19 77 14.966 4.9202$e-7
16 18 78 17.275 t G.79166e-$
31 15 17 ' 79 ' 19.749 7.98693e-9
32 14 ' 16 80 i 22.389 7.95573e-10
i 13 15 $1 25.195 i 6.6669$e-11
33
34 12 14 82 ' 2$.166 ~ 4.66306e-12
11 13 i $3 I 31.302 2,69671e-13
3G 1D ~ 12 ~ 84 i 34.604 1.27523e-14
37 9 11 85 3$.072 4.$6574e-16
61
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Table 3. (Continued}
Cell Longitudinal i
Frequencies
A b c d ~ AssociatianProbability
Score
38 8 10 ~ 86 41.705 i 1.47402e-17
39 ' 7 ' 9 87 45.504 3.47543e-19
40 6 8 ' 88 i 49.468 6.22023e-21
41 5 7 89 53.597 8.18226e-23
42 4 6 90 i 57.893 7.57617e-25
43 3 5 91 62.353 ' 4.64677e-27
44 2 ~ 4 92 66.979 ~ 1.72188e-29
45 1 3 ' 93 ~ 71.771 3.29152e-32
46 0 2 94 76.728 2.28366e-35
Table 4 is the same as Table 2 except that Table 4 portrays values of summary
strength of longitudinal association measures rather than summary longitudinal
association scores. The strength measure used in Table ~ is the measure
labeled SD
in Section 4.1.6 of the parent application. Section 4.1.6 of the parent
application
also describes computation of values of the strength of longitudinal
association
measures.
Table 4 portrays a different AGP than that portrayed in Table 2 in that the
two
AGPs use different measures of longitudinal association or temporal
contingency.
G2
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Table 4. A portrayal of the AGP for the hormone data that uses values of a
summary strength of longitudinal association measure.
Summary
of the
Delay
= 0 Specific
Portion
of the
ACP
Independenti Dependent
Variable
Variable GnRI-I P-LH J-LH P-FSH J-FSH
GnRH i .938 .587 I .782 ~ .297
i P-LH ~ .938 i .60D .912 i -.190
J-LH .456 ,534 i .356 .243
P-FSH .758 ' .912 .471 ' .207
~ J-FSH ' .215 -.191 ' .237 .148
i
Summary
of the
Delay
= 1 Specific
Portion
of the
ACP
' GnRH ' .911 .660 i .596 .482
I P-LH ~ .797 I .673 .587 .360
J-LH -.396 .446 -.346 ~ .243
'
' P-FSH .630 ~ .635 .608 ' .307
~ J-FSH ~ -.234 -.243 i .283 ~ -.210
~ Summary
of the
Delay
= 2 Specific
Portion
of the
ACP i
GnRH .523 .658 .353 .427 i
P-LH ~ -.469 I .645 t .356 ,364
I
J-LH -.396 .397 -.351 .204
P-FSH -.438 ' -.416 .565 ' .40D
J-FSH i -.253 i -.301 ~ -.182 -.270
~ Summary
the Entire
AGP
' GnRH .938 .660 i .782 .427 ,
P-LH .938 .673 .912 .364
J-LH .456 ' .534 .356 I .243
I P-FSH .758 .912 .608 .400 i
J-FSH i -.253 -.301 ~ .283 -.270 ~
Table 5 demonstrates a statistical analysis of a summary measure of strength
of
longitudinal association. The values in Table 5 axe For four associations and
sip
ewes. Labels suoh as "GnRH to P-LH" indicate that GnRH is functioning as the
independent variable and P-LH is functioning as the dependent variable. Table
5
includes the value of the first strength of longitudinal association measure
in Table
4. It is for Ewe 3 and is shown in bold in Table 5.
Nate that in Table 5, statistical tests are being performed on values of a
summary
strength of longitudinal association measure rather than measures of the
hornlones
themselves. The statistical test for any particular association is based on
only one
63
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
value for each ewe. The null hypothesis is that there is no longitudinal
association
between variables, that the mean score equals zero. The statistical tests are
single
group t-tests on the means. All p-values are for two-tailed Tests.
The values for individual ewes in Table 5 are conclusions, generalized across
up to
143 repeated measurements of each variable for each ewe, about cerhain
longitudinal associations fox each ewe. The means and standard deviations
(5.D.)
describe the associations for a group of six ewes. Assuming thafi the group is
a
representative sample of some population of ewes, the results of the
statistical tests
apply most directly to a collective entity, namely, the population of ewes
rather
than any particular ewe.
Table 5. Statistical analyses of values of a summary strength of longitudinal
association measure for 6 ewes.
Ewe ' T
Association1 2 ~ 3 5 6 ~ ~ ~ p
10 Mean S,D. ~
GnRH to 0.8790.881T 0.938' 0.734~ T 0.070<,0001
P-LH 0.878 0.8260,856
GnRH to 0.7700,6470.660 1.0000,6840,8540.7690,138<.0001
J-LH
GnRH to 0.5930,6520,782 O,G690.489' 0.5720.185.0006
P-FSH 0.251
GnRH to 0,4500.3750,427 0,871~ T T 0.259,0019
J-FSH 0.6711.0000.632
Figures 2 and 3 also are based on the hormone data. These figures demonstrate
the
use of ACPs to evaluate the temporal criterion of causal and other predictive
relationships. Both figures show average values of a summary strength of
longitudinal association measure in which the averages are obtained across the
six
ewes identified in Table 5. The group average value of a summary strength of
longitudinal association measure is shown as functions of the analysis
parameter
called delay.
Figure 2 portrays associations involving Gn RH and P-LH in which bofh hormone
measures function as both independent and dependent variables. The Icey that
identifies lines in Figure 2 includes reference to "Extensive" and "Limited."
These
labels refer to scoring options used in two scoring protocols. Scores based on
the
protocol labeled "Extensive" were the same as those described in paragraphs 3
and
6-4
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4 of this section except for additional values of delay (a total of 7 values
of delay, 0
through 6). In contrast, scores labeled "Limited" are based on a protocol that
used
default values for episode length, episode criterion, and persistence.
Figure 2 helps validate a new methodology (tVIQALA and AC.Ps) by coniln~ning
that the new methodology reveals a known relationship - namely that GnRH
secretion largely controls P-LH secretion. First, Figure 2 shows char patterns
in
average strength of longitudinal associations between GnRH and P-LH as a
function of delay. The most extreme values are positive and are for delay
equals
zero. Furthermore, the curves for the P-LH to GnRH interactions appear to be
shifted to the left relative to the corresponding GnRH to P-LH interactions.
Together, these results suggest that GnRH secretion tends to elicit P-LH
secretion
in a manner that is relatively rapid compared to the temporal resolution of
these
data-the temporal resolution being 5 minutes.
Results for delay equals 4 are among the most extreme negative average summary
strength of longitudinal association measure values shown in Figure 2. This
indicates that high levels ofthe independent variable are associated with low
levels
of the dependent variable about 20 minutes latter. The interactions portrayed
in
Figure 2 tend to be periodic.
Figure 2 also shows that values of the average summary strength of
longitudinal
association measure that are obtained with the "extensive" scoring protocol
options
are morn extreme (closer to the maximum values of plus or minus 1) than
corresponding values obtained with the "limited" options. This suggests that
the
opfional levels of the analysis parameters called episode length, episode
criterion,
and persistence do account for additional systematic variation in the data.
Figure 3 was obtained in fihe same manner as Figure 2 except that luteinizing
;0 hormone was measured in jugular blood rather than portal blood. High levels
of
GnRH are associated most strongly with high levels of J-LH about 10 minutes
later
as these variables were measured in these investigations. As indicated by the
difference between the "extensive" and the "limited" options for values of the
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
average summary strength of longitudinal association measure for the GnRH to J
LH interaction at delay zero; episode length, episode criterion, and
persistence can
tend to blur results for delay. However, these additional analysis parameters
do
tend to account for more systematic variation in the data as indicated by more
extreme positive and negative values.
Although Tables 2 through 5 and Figures 2 and 3 portray interactions invalving
only five hormone variables, these examples illustrate how this invention
could be
used to explore, discover, and confirm thousands or millions of interactions
involving hundreds or thousands of products of gene expression including
proteins.
Many of the computational procedures could be automated. As such, this
invention
can be an important and valuable contribution to bioinfonnatics. Furthermore,
this
invention can be applied to objects of investigation that are of other types
as
illustrated with Table 6.
Table 6 is an ACP that portrays internal control of the United States economy.
The
action involves ten variables used to obtain values of the Index of Leading
Economic Indicators, one variable that is a measure of the Gross Domestic
Product
of the United States economy, and one variable that is fhe Index of Leading
Economic Indicators. The data were obtained from The Conference Board and are
for 169 consecutive quarters ending recently.
The scoring protocol for Table 6 involves using residuals from second order
polynomial regression of each variable on measurement occasion and 12
dichotomous series to represent levels of each variable. Table 6 is for delay
equals
2. Default values were used for episode length, episode criterion, and
persistence.
Table 6 portrays summary strength of longitudinal association scores.
GG
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
IL-YoOOv' O ~OM ~~1'oOl~I(~
l0 ~ I ' O S~
i
Cs-7O IMI Wit'hd-Ic~10o,OO ooI ,
.--~ I ' I ' ~
I ~ v
lad'NM M ~~--~M M MCVV
I 1
.
O
O
I SCo0M~ ~ '~~'M N O~L--iCOI O~I U U
h COM ~ ~h'M'V~O\l010 N ~ ~ U vi
W h ~'~~ ~ ON .-,,~,--yy M U 1~-~ ''b
I ' '
~
I ~ I, i n 'I ~ ~, ~'. ,
I ' cd
I ' 3
I d Ir. ,~'OoyI oo~.,, ~ r,~ ~ I o
~ c-, o~ i I IC
I
f'OO~O~. Vl~OO\~ 0000 ~ ~fl[/] N
I M '-a' M .--~M 1 .-'0 -m-'~ ~j
~, 1 ccj +-' 'C~
i'Y~C n I~n ~ n I~ , I
I I' I I I
. I'
I I I
I I
~
I U ~ I~OO N 'NI N h IV1V1M IP
~, ~ h hM I -~O .-.VY .-..-a~ ,k' 'd
N I
M ~N IN.-iI,~N N I~~'~n~ II ,
. I
I I
I I
I
~
U I ~ h NO M '~M p~I NN O ,...O bn
' I I '
U .S
U p M O~ O O~h INI O~'0010,
'~Lr-YN ~O'O'\ON '~"aIr-.M y.~ ci
U
in
~ ~ ~ ~
N
~ S-i
% I U 4-1
~ O 'S""
C I ~ ~.J
7 ~
i i-
b~d'IOVM V~O01 ~n~O o000-~ c/~
~ ~ ~ ~ N
~ ~ O '~~~ 00~~0 ' N ~~ M M ~ ' ~ c~
1 I I I I~,INU .~"_, i-i I
~'G ~ _
c O
I I I I
I I '
U
N I I I U
I I I I I I I O tti
c~pp00'-' l~M IOVO~MO ~ON Q7 N
>-'~ U 'ctO~ ~ ~' ~ O .-.I~'O'o0I ~ p
=0 d''-'N I I'-~ N lDN. ~'C'7
. l0
I vy.
, ~
, , O
U ~ cad
U
II I ~
IL I I ~., S-,
N +~
IN\O h 'O~ INIl~10.~N I
I
"
',
~ ~ M~ O I~ ~ II
'- Y S.,
N N
~ ~ . I: N ~ N . O-
. . . L
t~, I I I I C~ ~ 4o b
I
"' '
I y ~;
I N M M~ ~ II
~ 1~ 000 ~ ~O' ~ U
IUh M~ \Od'M O M~ ~~1r7
~ ~ Id' I IO I
I N V N IM
M NI M 1 I~ N
I I I I I I a-.~ ,-~l
I 0
.~ N T
~ N
I I I I I ~
h 1 h 4
~
N Id-~O0 d'oor., N ~ . J,
I 'I ~' II
N ~O\O ~!1V1IcfV7(yVpp00M v3
~ ' ' ' I 7 '
Tj ~ (V..M M N.-aIM"...,Cue,O CV~ U ~ ,~
. '. . ~. . . , I . O S~
. .
~~I I I II I ' Y I I
~ I I
I I I I v d.o
I
' ~'
U ~ N 'Ovd' '00I~I,~00Ov .
Ir-._
I I ~ ~ Vj ~
c~C 10 N ~ ~O r O N~ c~,,~ ~ ~ c~ SZ. O
' ' I
~
I ' n I II . ~ i-
' I -n V
O
I I I I II O III ~ rA
~ I
IP 01~ M .-wryr-.01l~~ 00l~
-
I
I ~ ~Id I~iiI~ I~I~ ~
1 M
~
I , - IM N M S... ~ ~ U
~r- L. , .
~J .
I I II 40~ ~ cat '~
I I
1 ~'
,
I ~ I I I v~
by
' I ~
.,~ I I U
I I I I I O ~ O
"'' O
bD
"C7~ I5~. ~ '~i~. ~ J, ~. cd ~
_ ~ c~ I Is-s.:
I ~
(.~~ "~7I ~IYk , J U U 'd .r-'
U N'~ U csW'W'
O
4-r O N c~ ~ ;-. O
I'~ ~ ~ I~I~'IUIr-~~ O
I
I I I ..O ~
S7 ~', ~ ~
~..,
I I I I t
.
Q ~ O ~ O
I I
V'7
G, r~
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
Interpretation of Table G will be illustrated with the dependent variable
column labeled
GDP. Independent variables labeled MfCG, Stock, MfCap, and CsExp are the most
powerful predictars of GDP, given these data and the particular scoring
protocol used
for Table 6. The independent variables labeled Unemp and Rate are negatively
associated with GDP. As expected, L,EI is positively predictive of GDP.
While the best mode for carrying out the invention has been described in
detail, those
familiar with the art to which this invention relates will recognize various
alternative
designs and embodiments for practicing the invention as defined by the
following
claims.
G
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
APPENDIX
Outline of Application
1. BACKGROUND OF THE INVENTION
1.1. Technical Field
1.2. Description o~Related Art
1.2.1. The Need to Measure Interactions that are Temporal Contingencies
1.2.2. Specific Problems Involved in the Prior Art
1.2.2.1. Problems Involving Individuality
1.2.2.2. Problems Involving Complexity
1.2.2.3. Problems Involving Nonlinearity
1.2.2.4. Problems Involving Comprehensiveness
1.2.2.5. Problems Involving Detail
1.2.2.6. Need to Investigate All Five Types of Problem as a Set
1.2.3. Citations
2. BRIEF SUMMARY OF THE INVENTION
2.1. Structure of ACPs
2.2. Functions oi'ACPs
2.3. Haw do ACPs Help Address Limitations of the Statistical Method?
2.3.1. MQALA and the Statistical Method Are Best Suited to Analyze Distinct
Types of Data (Evidence)
2.3.2. MQALA and the Statistical Method Have Distinct Objectives
2.3.3. MQALA and the Statistical Method Use Distinct Computational
Procedures
2.3.4. MQALA and the Statistical Method Are Best Suited for Distinct Types
of Entities
2.4. MQALA Helps Address Problems Described in Section 1.2.2.
2.4.1. ACPs Help Address Problems Involving Individuality
2.4.2. ACPs Help Address Problems Involving Complexity
2.4.3. ACPs Help Address Problems Involving Nanlinearity
2.4.4. ACPs Help Address Problems Involving Comprehensive Investigations
2.4.5. ACPs Help Address Problems Involving Detailed Investigations
2.4.6. Addressing the Need to Investigate All Five Types of Problem as a Set
3. BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWI ~ ~G
4. DETAILED DESCRIPTION OF THE INVENTION
4.1. Objects o~ Investigation and Their Environments
4.1.1. Organisms Including Persons
4.1.2. Portions olOrganisms
4.1.3. Economies and Investment Markets
4.1,4. Machines, Processes, and Other Man Made Systems
4.1.5. Systems Consisting oI'Two or More Individuals
4.1.6. Populations Investigated as Wholes
4.1.7. Nested Systems
4.1.8. Additional Types oC Systems
G9
CA 02425871 2003-04-09
WO 02/31604 PCT/USO1/31414
4.2. Actions
4.2.1. Movement
4.2.2. Physical and Electromagnetic Action
4.2.3. Chemical ar Biochemical Action
4.2.4. Biological Action
4.2.5. Emotional, Mental, and Behavioral Action
4.2.6. Social and Economic Action
4.2.7. Action Measurement Technologies
4.2.7.1. Biochemical Measurement Technologies
4.2.7.2. Functional Imaging
4.2.7.3. MicroElectroMechanical Systems (MEMS)
4.2.7.4. Instrumentation for Psychophysios and Psychometrics
4.2.7.5. Performance Measures
4.2.7.6. Rating Scales and Surveys
4.2.8. Sciences and Disciplines
4.3. Computational Features
4.4. Measures Portrayed in ACPs
4.5. Portrayal of ACPs
4.6. Databases that Include ACPs
4.7. Analyses ofACPs with Statistics or Other Quantitative Methods
4.8. Uses of ACPs
4.8.1. Use of ACPs to Measure Internal Control
4.8.2. Use of ACPs to Measure Responses to Environments
4.8.3. Use of ACPs to Measure Actions on Environments
4.8.4. Use ofACPs to Fingerprint Individuals
4.8.5. Use of ACPs to Portray Test Results
4.8.6. Use of ACPs to Identify Predictors of Disorder
4.8.7. Use of ACPs to Identify Predictors ofDifferential Response
4.8.8. Use of ACPs to Measure Interactions Involving Different Types of
Action
4.8.9. Use ofACPs to Help Distinguish Causal from Non-causal Associations
4.8.9.1. Exploring for Causal Relationships
4.8.9.2. Confirming Causal Relationships
4.8.10. Use of ACPs for Data Mining
4.8.11. Use of ACPs for Nested Systems
4.8.12. Use ofACPs to Distinguish Episodes of Action
4.8.13. Use ofACPs for Model Development and Testing
4.8.14. Use of ACPs to Draw Generalized Conclusions and to Make Predictions
4.8.15. Use ofACPs to Make Scientific Discoveries
4.8.16. Use of ACPs to Guide Decision-Making
4.9. Examples ofACPs
S. Appendix
Claims
