Methods of identifying subjects for inclusion and / or exclusion in a clinical trial

Abstract

Disclosed herein are methods of selecting a candidate subject based on predicted response to an experimental treatment or therapeutic agent, and related methods of identifying subjects eligible to participate in a clinical trial that may be likely to respond to placebo. The methods disclosed herein are useful in the identification and / or selection of subjects for inclusion or exclusion from the clinical trial.

Description

METHODS OF IDENTIFYING SUBJECTS FOR INCLUSION AND / OR EXCLUSION IN A CLINICAL TRIALRELATED APPLICATIONS PARAGRAPH

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 533,888, filed August 21, 2023. The entire teachings of the above application are incorporated herein by reference.FIELD OF THE INVENTION

[0002] The present inventions relate to methods of identifying subjects for inclusion and / or exclusion in a clinical trial based upon the subject’s determined likelihood of responding to a drug and / or placebo. In particular, the present invention relates to the development, training and application of a set of algorithms that allow for the quantitative selection of individual subjects for inclusion in a central nervous system (CNS) clinical trial in relation to a specific neural target.BACKGROUND OF THE INVENTION

[0003] Generally, subjects participating in a clinical trial for an experimental treatment are selected based upon inclusion criteria, which help to define subjects who match the population of patients that the trial aims to study, and exclusion criteria, which exclude subjects for whom the experimental treatment may be unsafe or ineffective. These inclusion and exclusion criteria are established based upon several factors, including, for example, the indication or symptom that the experimental treatment (e.g., a drug) is intended to treat or address, the outcomes of interest, regulatory guidelines and prior evidence, such as available safety and efficacy data obtained from preclinical studies and prior clinical trials.

[0004] Such traditional sources of information help to define the target population for the experimental treatment but may experience and be subject to several shortcomings. For example, there is a lack of available technologies capable of identifying individual subjects who are likely to respond to the experimental treatment in a clinical trial based upon their predicted likelihood of response to a specific neural target. There are currently no quantitative methods that use functional neural information for selecting or excluding subjects from participating in a clinical trial based upon their predicted probability of responding to the experimental treatment.Description of the Related Art

[0005] CNS drug development faces a high rate of failure, in part because there is massive heterogeneity within CNS diagnostic categories. For example, patients with the same diagnosis may not always share the same underlying pathology, leading to a high proportion of non-responders or partial responders to administered treatments and medications. These challenges are exacerbated because many clinical trials for new experimental treatments of CNS conditions are designed to enroll subjects meeting specific behavioral diagnostic criteria without any way to differentiate between potential subgroupings or sub-populations of subjects who may be more or less likely to respond to the experimental treatment. There are currently no methods for optimizing the enrollment of subjects in a clinical trial based on such subjects’ predicted likelihood of response to an experimental treatment (e.g., a drug).

[0006] There is a growing effort to include neurobiologic al and / or neuroimaging markers in the development and ultimately prescription of novel CNS medications. For example, lecanemab, which is prescribed to slow the progression of Alzheimer’s disease, requires confirmation of the presence of amyloid beta plaques in the patient’s brain through a sample of cerebrospinal fluid or a positron emission tomography (PET) scan prior to prescription, as well as regular magnetic resonance imaging (MRI) scans of the patient for safety monitoring during treatment. However, the approach of requiring a neural scan or lumbar puncture for every potential patient as a prerequisite for treatment is costly, invasive and impractical.

[0007] It is important to gain an understanding of how diverse symptom presentations relate to biological mechanisms and circuitry in the human brain. This has been greatly aided by noninvasive imaging techniques, such as MRI, which has mapped neural functioningrelated to a wide range of behaviors, including psychiatric symptoms (e.g., psychosis and anhedonia), deficits in executive function and cognition (e.g., verbal reasoning and working memory), and developmental processes. These neuroimaging approaches have further been used along with behavioral phenotype data in “unsupervised” machine learning models to discover patterns of neural and behavioral relationships without explicit guidance on what those relationships should be, which can provide insight into the circuits underlying CNS disorder symptoms (see, Ji, et al., Elife 2021, Jul 20: 10:e66968).

[0008] United States Patent Application Publication No. 2021 / 0005306 describes computing neuro-behavioral geometries consisting of neural features statistically mapped to behavioral features defined in a group of individuals exhibiting a variety of symptoms. This publication describes methods of determining a therapeutic for a patient with a neural disorder via such a neuro-behavioral geometry, consisting of first deriving neural maps that are associated with unsupervised data-reduced behavioral dimensions in a group of patients along that CNS disorder spectrum, and then identifying candidate therapeutic targets that are associated with the derived neural maps. Thus, a potential therapeutic for a new patient can be identified based on the patient’ s location along one or more dimensions in the neuro- behavioral geometry. However, this publication does not describe methods pertaining to “supervised” approaches (i.e., if the algorithm for detecting neuro-behavioral relationships is given the constraint that the neural maps must be maximized in similarity to a specific target map). In other words, this publication does not describe how to determine a set of behaviors (or individuals with those behaviors) that maximally align to a neural map for a given target.

[0009] In addition to noninvasive neuroimaging, neural circuits can be characterized by information at the microscale level, such as by using expression levels of different genes in the brain. Recent efforts have measured spatially comprehensive atlases of the neural transcriptome using post-mortem human brains (see, Hawrylycz, et al., 2012, Nature 489: 391-99). These transcriptomic data have been combined with state-of-the-art imaging methods to produce gene expression maps that leverage surface-based cortical topographies and that can be related to brain activity associated with a behavioral phenotype of interest (see, Burt et al., 2018, Nature Neuroscience 21: 1251-59). United States Patent No. 11,791,016 describes approaches that can confirm the association between a neural phenotype map representing behaviorally-relevant brain activity, and a gene expression “target” map that is associated with a drug target. Specifically, this association between the two maps can be quantified using a similarity score for each phenotype-gene pair, and genetarget maps that are highly similar to a neural phenotype map for a specific symptom or disorder can in turn be used to inform potential treatment targets for patients exhibiting the phenotype. However, the methods described in this published patent require that both the gene expression target maps and patient neural phenotype maps be previously independently characterized. Thus, they can be used to confirm the association between a set of phenotypes and a set of gene expression maps associated with therapeutic targets but cannot be used to compute the phenotypes that are associated with a given therapeutic target map. Moreover, these disclosed methods do not provide a way to derive the relative importance of specific symptoms / behavioral measures with respect to a neural target map of interest, such as a gene expression map (i.e., the methods disclosed in the published patent cannot, given a specific neural target map, identify which symptoms and / or behaviors are most relevant to the neural circuits that are most similar to the target map).

[0010] Needed are novel methods and strategies of aligning a set of behavioral features and their associated patterns of brain activity with a specified neural target map, and for identifying subjects for inclusion and / or exclusion in a clinical trial based upon the subject’s determined likelihood of responding to a drug and / or placebo. Also needed are novel methods and strategies for optimizing enrollment of subjects in a clinical trial. The present invention is directed toward further solutions to address these needs, in addition to having other desirable characteristics.SUMMARY

[0011] Disclosed herein are novel methods of facilitating the identification of subjects that are likely to respond to an experimental treatment (e.g., a drug) so that such subjects are included in the clinical trial. In certain embodiments, the methods disclosed herein may be used to facilitate the identification of subjects that are likely to respond to a placebo so that such subjects are excluded from the clinical trial. In contrast to traditional strategies, the methods disclosed herein provide a set of behavioral features that reflect patterns of brain activity that in turn align with a specified neural target map. Also, in contrast to traditional strategies, the methods disclosed herein only require assessments of a candidate subject’s behavioral features to predict their likelihood of responding to an experimental treatment.These behavioral features can then be used as enrollment criteria to enrich clinical trials with subjects that exhibit behaviors which are most likely to be associated with the neural target.

[0012] In certain embodiments, the inventions disclosed herein concern a method of selecting a subject as a candidate subject in a clinical trial based on their predicted response to an experimental treatment or therapeutic agent, such method comprising: (a) a step of retrieving, by at least one processor of a computing device, a neural supervision target benchmark map associated with the experimental treatment or therapeutic agent, wherein said neural supervision target benchmark map comprises neural target data collected from one or more individuals, the neural target data being associated with the experimental treatment or therapeutic agent, and wherein said neural supervision target benchmark map further comprises two-dimensional surfaces and subcortical volumetric structures in which the two- dimensional surfaces and the subcortical volumetric structures comprise one or more numerical values assigned to specific brain locations represented by the two-dimensional surfaces and the subcortical volumetric structures; (b) a step of retrieving, by the at least one processor of the computing device, a reference neuro-behavioral association dataset, wherein said reference neuro-behavioral association dataset comprises neural target data collected from one or more individuals, the neural target data including neural features and symptom features related to (or in certain instances, corresponding to) mental health or cognitive status; (c) a step of determining, by the at least one processor of the computing device, a symptom- to- alignment predictor function reflecting a relationship for the alignment between the neural supervision target benchmark map and the reference neuro-behavioral association dataset, wherein said determining comprises computing, via the at least one processor of the computing device, neural alignment scores for said reference neuro-behavioral association dataset relative to said neural supervision target benchmark map, wherein a high absolute neural alignment score of said reference neuro-behavioral association dataset with said neural supervision target benchmark map is indicative of a statistical correspondence between the neural features in said reference neuro-behavioral dataset and the neural target data in said neural supervision target benchmark map, wherein said neural alignment scores are used to compute the symptom-to-alignment predictor function that is indicative of a statistical correspondence between symptom features in the reference neuro-behavioral dataset and said neural alignment scores; and (d) a step of determining, by the at least one processor of the computing device, a neural alignment score of said subject, wherein said determining comprises assessing the symptom-to-alignment predictor function relative to said subject’s symptoms and / or assessing the neural supervision target benchmark map relative to said subject’s neural feature map; wherein a high neural alignment score is indicative of a highprobability of alignment with said neural supervision target benchmark map; and wherein based on said high probability of alignment, said subject is given a quantitative likelihood of response to said experimental treatment or therapeutic agent. It is important to note that the neural supervision target benchmark map is specified a priori and acts as a constraint for the subsequent steps described here, as the goal of this method is to select subjects with the maximal predicted neural similarity to the neural supervision target benchmark map (e.g., subjects whose neural features map most highly resemble the selected neural supervision target benchmark map).

[0013] The methods set forth in the preceding paragraphs and further disclosed herein stand in contrast to, and are readily distinguishable from the “unsupervised” methods described in United States Patent Application Publication No. 2021 / 0005306, the disclosure of which is incorporated herein by reference, which prioritize first defining the relationships between dimensions of symptom / behavior variation and neural features derived from neural data collected in a population of individuals to derive neuro-behavioral maps, and subsequently identifying candidate therapeutic targets that are associated with the derived neural-behavioral maps. The supervised methods described in the instant application select subjects maximized in similarity to a pre-specified target benchmark map, whereas the unsupervised methods in United States Patent Application Publication No. 2021 / 0005306 select target benchmark maps based on their similarity to symptom-relevant neural- behavioral maps.

[0014] In certain embodiments of the foregoing methods a high quantitative likelihood of response is indicative of the subject being selected as a candidate subject predicted to respond to said experimental treatment or therapeutic agent. Conversely, in certain embodiments of the foregoing methods a low quantitative likelihood of response is indicative of the subject being selected as a candidate subject that is not predicted to respond to said experimental treatment or therapeutic agent.

[0015] In certain embodiments, the foregoing methods may further comprise a step of selecting said candidate subject as likely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial. For example, in certain embodiments, a candidate subject with maximal predicted neural similarity to the neural supervision target benchmark map is selected for inclusion in the clinical trial and is randomized to receive the experimental treatment (e.g., a drug) or a placebo. Alternatively, in other embodiments, the foregoing methods may further comprise a step of excluding saidcandidate subject as unlikely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial. For example, in certain embodiments, a subject with a low predicted neural similarity to the neural supervision target benchmark map is excluded from the clinical trial. In some embodiments, the methods disclosed herein may further comprise a step of administering an experimental treatment or therapeutic agent to the subject selected as a candidate subject predicted to respond to said experimental treatment or therapeutic agent. In some embodiments of the foregoing methods, if the subject is selected as a candidate subject not predicted to respond to said experimental treatment or therapeutic agent, said experimental treatment or therapeutic agent is not administered to said subject.

[0016] In some embodiments of the methods disclosed herein, the neural supervision target benchmark map comprises a brain- wide PET map of receptor occupancy for the therapeutic agent. In certain aspects, the neural supervision target benchmark map comprises a pharmacological map associated with one or more receptor targets. In yet other embodiments, the neural supervision target benchmark map comprises a gene expression map associated with one or more gene expression targets. In other embodiments, the neural supervision target benchmark map comprises a previously computed neuro-behavioral variation map associated with one or more symptoms and / or indications. In other embodiments, the neural supervision target benchmark map comprises a task-evoked neural map associated with one or more functions.

[0017] In certain aspects, the neural supervision target benchmark map is in a space where the subject’s left and right brain hemispheres are represented as surfaces and the subcortex is represented as a volume. In certain embodiments, the neural supervision target benchmark map and the reference neuro-behavioral association dataset are in the same space. In certain embodiments, the neural supervision target benchmark map and the reference neuro-behavioral association dataset are parcellated according to an atlas of functionally- defined regions. Some examples of such regions include the primary visual cortex, fusiform face area, medial prefrontal cortex, and the perisylvian language area. These regions are defined not only based on neuroanatomical and microstructural properties but also on functional measures such as functional MRI.

[0018] Alternatively, in other embodiments, the neural supervision target benchmark map is at the level of networks, areas, or individual vertices or voxels.

[0019] In a particular embodiment, the symptom-to-alignment predictor function is a linear regression model that produces a set of linear weightings that predict neural alignmentscores from symptom measures. In other embodiments, the symptom-to-alignment predictor function is a non-linear mapping that predicts neural alignment scores from symptom measures.

[0020] Also disclosed herein are methods of selecting a subject predicted to respond to a placebo, the methods comprising: (a) a step of retrieving, by at least one processor of a computing device, a placebo-relevant neural supervision target benchmark map, wherein said placebo-relevant neural supervision target benchmark map comprises neural target data collected from one or more individuals that participated in a clinical trial comprising a placebo arm; and wherein said placebo-relevant neural supervision target benchmark map comprises two-dimensional surfaces and subcortical volumetric structures in which the two- dimensional surfaces and the subcortical volumetric structures comprise one or more numerical values assigned to specific brain locations represented by the two-dimensional surfaces and the subcortical volumetric structures; (b) a step of retrieving, by the at least one processor of the computing device, a reference neuro-behavioral association dataset, wherein said reference neuro-behavioral association dataset comprises neural target efficacy data collected from the one or more individuals that participated in the clinical trial and were randomized to the placebo arm, the neural target efficacy data including neural features and symptom features; (c) a step of determining, by the at least one processor of the computing device, a symptom-to-alignment predictor function reflecting a relationship for the alignment between the placebo-relevant neural supervision target benchmark map and the reference neuro-behavioral association dataset, wherein said determining comprises computing, via the at least one processor of the computing device, neural alignment scores for said reference neuro-behavioral association dataset relative to said placebo-relevant neural supervision target benchmark map, wherein a high absolute neural alignment score of said reference neuro-behavioral association dataset with said placebo-relevant neural supervision target benchmark map is indicative of a statistical correspondence between the neural target efficacy data in the reference neuro-behavioral association dataset and the neural target data in the placebo-relevant neural supervision target benchmark map, wherein said neural alignment scores are used to compute the symptom-to-alignment predictor function that is indicative of a statistical correspondence between symptom features in the reference neuro- behavioral dataset and said neural alignment scores; and (d) a step of determining, by the at least one processor of the computing device, a neural alignment score of said subject, wherein said determining comprises assessing the symptom-to-alignment predictor function relative tosaid subject’s symptoms and / or assessing said placebo-relevant neural supervision target benchmark map relative to said subject’s neural data, wherein a high neural alignment score is indicative of a high probability of alignment with said placebo-relevant neural supervision target benchmark map, and wherein based on said high probability of alignment, said subject is given a quantitative likelihood of response to placebo.

[0021] In certain embodiments, subjects identified as having a high neural alignment score have a high likelihood of being a placebo responder, and subjects having a low neural alignment score have a low likelihood of being a placebo responder.

[0022] In certain embodiments, the foregoing method further comprises a step of excluding said subject as a likely placebo responder prior to randomization in a clinical trial.100231 In certain aspects of the present invention, the placebo-relevant neural supervision target benchmark map comprises a brain- wide PET map of receptor occupancy associated with one or more receptor targets. Alternatively, in other embodiments, the placebo-relevant neural supervision target benchmark map comprises a pharmacological map associated with one or more receptor targets. In yet other embodiments, the placebo-relevant neural supervision target benchmark map comprises a gene expression map associated with one or more gene expression targets. In certain embodiments, the placebo-relevant neural supervision target benchmark map comprises a previously computed neuro-behavioral variation map associated with one or more symptoms and / or indications. In yet other embodiments, the placebo-relevant neural supervision target benchmark map comprises a task-evoked neural map associated with one or more functions. In certain embodiments of the foregoing methods, the placebo-relevant neural supervision target benchmark map is in a space where the subject’s left and right brain hemispheres are represented as surfaces and the subcortex is represented as a volume.

[0024] In certain embodiments, the method for selecting subjects based on predicted placebo response is used in conjunction with the above method for selecting subjects based on predicted response to a therapeutic agent.

[0025] In certain embodiments disclosed herein, the placebo-relevant neural supervision target benchmark map and the reference neuro-behavioral association dataset are in the same space. In certain aspects, the placebo-relevant neural supervision target benchmark map and the reference neuro-behavioral association dataset are parcellated according to an atlas of functionally-defined regions. Examples of these regions are set forth supra and include, forexample, the primary visual cortex, fusiform face area, medial prefrontal cortex and the perisylvian language area.

[0026] In yet other embodiments of the foregoing methods, the placebo-relevant neural supervision target benchmark map is at the level of networks, areas, or individual vertices or voxels.

[0027] In a particular embodiment of the methods disclosed herein, the symptom-to- alignment predictor function is a linear regression model that produces a set of linear weightings that predict neural alignment scores from symptom measures. In other embodiments, the symptom-to-alignment predictor function is a non-linear mapping that predicts neural alignment scores from symptom measures.100281 Also disclosed herein are methods of selecting a subject as a candidate subject based on their predicted response to an experimental treatment or a therapeutic agent, the method comprising: (a) a step of retrieving, by at least one processor of a computing device, a neural supervision target benchmark map, wherein said neural supervision target benchmark map comprises neural target efficacy data collected from one or more individuals that participated in a clinical trial of the experimental treatment or therapeutic agent, the clinical trial comprising a treatment arm and a placebo arm; and wherein said neural supervision target benchmark map comprises two-dimensional surfaces and subcortical volumetric structures in which the two-dimensional surfaces and the subcortical volumetric structures comprise one or more numerical values assigned to specific brain locations represented by the two-dimensional surfaces and the subcortical volumetric structures; (b) a step of retrieving, by the at least one processor of the computing device, a reference neuro- behavioral association dataset, wherein said reference neuro-behavioral association dataset comprises baseline clinical data collected from one or more individuals, the baseline clinical data including neural target efficacy data and symptom features related to (or in certain instances, corresponding to) mental health or cognitive status; (c) a step of determining, by the at least one processor of the computing device, a symptom-to-alignment predictor functions reflecting a relationship for the alignment between the neural supervision target benchmark map and the reference neuro-behavioral association dataset, wherein said determining comprises computing, via the at least one processor of the computing device, neural alignment scores for said reference neuro-behavioral association dataset relative to said neural supervision target benchmark map, wherein a high absolute neural alignment score of said reference neuro-behavioral association dataset with said neural supervisiontarget benchmark map is indicative of a statistical correspondence between the neural target efficacy data in the reference neuro-behavioral association dataset and the neural target efficacy data in the neural supervision target benchmark map, wherein said neural alignment scores are used to compute the symptom-to-alignment predictor function that is indicative of a statistical correspondence between symptom features in the reference neuro-behavioral dataset and said neural alignment scores; and (d) determining, by the at least one processor of the computing device, a neural alignment score of said subject, wherein said determining comprises assessing the symptom-to-alignment predictor function relative to said subject’s symptoms and / or assessing the neural supervision target benchmark map relative to said subject’s neural data; wherein a high neural alignment score is indicative of a high probability of alignment with said neural supervision target benchmark map, and wherein based on said high probability of alignment, said subject is given a quantitative likelihood of response to said experimental treatment or therapeutic agent.

[0029] In certain aspects of the methods described in the preceding paragraphs, a high quantitative likelihood of response is indicative of the subject being selected as a candidate subject predicted to respond to said experimental treatment or therapeutic agent, and a low quantitative likelihood of response is indicative of the subject being selected as a candidate subject that is not predicted to respond to said experimental treatment or therapeutic agent.

[0030] In certain embodiments, this method further comprises a step of selecting said candidate subject as likely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial. In other embodiments, such method further comprises a step of excluding said candidate subject as unlikely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial. In yet another embodiment, such method further comprises a step of administering said experimental treatment or therapeutic agent to said candidate subject predicted to respond to said experimental treatment or therapeutic agent. In some embodiments, if the subject is selected as a candidate subject that is not predicted to respond to said experimental treatment or therapeutic agent, said experimental treatment or therapeutic agent is not administered to said subject.

[0031] In certain embodiments, the methods for selecting subjects based on their predicted response to an experimental treatment or a therapeutic agent is used in conjunction with the method for selecting a subject predicted to respond to a placebo.

[0032] In some embodiments of the methods disclosed herein, the neural supervision target benchmark map comprises a brain-wide PET map of receptor occupancy for said therapeutic agent. In some embodiments, the neural supervision target benchmark map comprises a pharmacological map associated with one or more receptor targets. In yet other embodiments, the neural supervision target benchmark map comprises a gene expression map associated with one or more gene expression targets. In some aspects, the neural supervision target benchmark map comprises a previously computed neuro-behavioral variation map associated with one or more symptoms and / or indications. In some embodiments, the neural supervision target benchmark map comprises a task-evoked neural map associated with one or more functions. In yet another embodiment, the neural supervision target benchmark map is in a space where the subject’s left and right brain hemispheres are represented as surfaces and the subcortex is represented as a volume. In certain aspects, the neural supervision target benchmark map and the reference neurobehavioral association dataset are in the same space.

[0033] In certain embodiments, the neural supervision target benchmark map and the reference neurobehavioral association dataset are parcellated according to an atlas of functionally-defined regions, as defined above. Examples of such regions may include the primary visual cortex, fusiform face area, medial prefrontal cortex, and the perisylvian language area. In other embodiments, the neural supervision target benchmark map is at the level of networks, areas, or individual vertices or voxels.

[0034] In a particular embodiment of the methods disclosed herein, the symptom-to- alignment predictor function is a linear regression model that produces a set of linear weightings that predict neural alignment scores from symptom measures. In other embodiments, the symptom-to-alignment predictor function is a non-linear mapping that predicts neural alignment scores from symptom measures.

[0035] The above discussed, and many other features and attendant advantages of the present inventions will become better understood by reference to the following detailed description of the invention.BRIEF DESCRIPTION OF THE FIGURES

[0036] The foregoing and other objects, features and advantages will be apparent from the following description of particular embodiments of the disclosure, as illustrated in the accompanying drawings.

[0037] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawings will be provided by the Office upon request and payment of the necessary fee.

[0038] FIG. 1 depicts a diagram of the constituent parts and steps of one embodiment of the present inventions where a selection model can be deployed to enrich a clinical trial with candidate subjects based on their predicted response. As shown, the methods constitute a “model training” step which leverages existing data, and a “model application” step which can be used in new clinical trials to select candidate subjects based on their predicted response to the therapeutic agent, without the need for novel neural data collection.

[0039] FIG. 2 represents a depiction of the benefits of using the novel methods disclosed herein in a clinical trial. As shown, the use of such methods in the EMB ARC clinical trial (Trivedi, et al., J. Psychiatr. Res., 2016, 78:11-23) would have resulted in better power of the study, resulting in less time and lower cost.

[0040] FIGS. 3A-3B depict a representative use case of the novel methods disclosed herein for the purpose of enriching a clinical trial with candidate subjects that are likely responders to the therapeutic agent. As shown in FIG. 3A, a strong placebo response reduces the effect size and the lack of exclusion criteria for strong placebo responders weakens clinical trial success. In contrast, the novel methods disclosed herein and, in particular, the use of the methods to identify subjects with high predicted alignment scores for inclusion in the clinical trial, increases the effect size and reduces the number of subjects required in the clinical trial, as shown in FIG. 3B.

[0041] FIG. 4 presents a block diagram of a process for performing the computation framework relating to enriching a clinical trial with candidate subjects that are likely responders to a therapeutic agent. As depicted, the model may be trained based upon an existing dataset.

[0042] FIGS. 5A-5F depict representative analysis results of the novel methods disclosed herein for the purpose of enriching a clinical trial with candidate subjects that are likely responders to a therapeutic agent. The novel methods disclosed herein were applied to data from a clinical trial investigating the effects of sertraline in depression (Trivedi, et al., J. Psychiatr. Res., 2016, 78:11-23). FIG. 5A depicts a gene expression map of SLC29A4, a gene encoding a target for sertraline. The color scale of the gene expression map reflects the relative level of expression (Z-score) in each area of the brain, with yellow indicating a positive value (i.e., relatively high level of expression of the SLC29A4 gene), and bright blueareas indicating a negative value (i.e., a relatively low level of expression of the SLC29A4 gene). FIG. 5B depicts the symptom weights obtained from fitting the symptom-to- alignment predictor function to the clinical trial data. Here, the symptom- alignment-predictor function is a linear regression, represented by weights shown in FIG. 5B. FIG. 5C presents a legend for the symptoms represented in FIG. 5B. FIG. 5D depicts a histogram of the alignment scores, and the 20% of subjects with the highest alignment scores (i.e., those subjects that would be selected for inclusion in the clinical trial by the claimed method) are highlighted. FIG. 5E shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for inclusion in the clinical trial by the claimed method. FIG. 5F shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). FIG. 5F thus illustrates the sample size reduction enabled by the claimed methods.

[0043] FIGS. 6A-6B present a representative use case of the novel methods disclosed herein for the purpose of excluding likely placebo responders from a clinical trial. FIG. 6A depicts that the lack of exclusion criteria for strong placebo responders weakens clinical trial success. In particular, the lack of exclusion criteria in the clinical study for strong placebo responders reduces the effect size. As shown in FIG. 6B, the use of the methods disclosed herein reduces the observed placebo response in remaining trial participants and improves trial success. Excluding subjects with a high predicted placebo response from the clinical trial would have resulted in an increase in the effect size and a reduction in the number of subjects required in the clinical trial, as shown in FIG. 6B.

[0044] FIG. 7 presents a block diagram of a process for performing the computation framework relating to excluding candidate subjects with high predicted placebo response from a clinical trial.

[0045] FIGS. 8A-8E depict representative analysis results of the novel methods disclosed herein for the purpose of excluding candidate subjects with high predicted placebo response from a clinical trial. FIG. 8A shows the symptom weights obtained from fitting the symptom-to-alignment predictor function to the trial data, and FIG. 8B provides a legend for the symptoms represented in FIG. 8A. FIG. 8C shows a histogram of the alignment scores; 80% of subjects with the lowest alignment scores are selected for exclusion by the claimedmethod, the remaining subjects (i.e., those that are selected for inclusion) are highlighted. FIG. 8D shows the anticipated effect size as a function of the percentage of candidate subjects with the highest alignment scores that would be selected for exclusion by the claimed method. Effect sizes are calculated as the inverse-variance weighted average Cohen’s d obtained in the test-sets of a 10-times repeated 2-fold cross-validation of the model. Negative values represent a reduction in symptom score such that lower effect sizes are better. The black dashed line denotes the effect size when all subjects would be selected for inclusion (i.e., without application of the claimed methods). FIG. 8E shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). FIG. 8E thus illustrates the required sample size reduction enabled by the claimed method.

[0046] FIGS. 9A-9F present representative analysis results of the novel methods disclosed herein for the purpose of increasing the effect size in a clinical trial by both enriching it with likely responders and excluding subjects with high predicted placebo response. FIG. 9A shows the gene expression map of SLC29A4, a gene encoding a target for sertraline, which gene expression map was used as the target map for the enrichment model. FIG. 9B shows the symptom weights obtained from fitting the enrichment model, here a linear regression, to the trial data, and FIG. 9C shows the symptom weights obtained from fitting the placebo model, here a linear regression, to the trial data. FIG. 9D provides a legend for the symptoms represented in FIG. 9B and FIG. 9C. FIG. 9E shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for inclusion in the enrichment model and selected for exclusion in the placebo model, and the black dashed line denotes the effect size when all subjects would be selected for inclusion (i.e., without application of the claimed method). FIG. 9F shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate, and the black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). As shown in FIG. 9F, the sample size reduction is enabled by the claimed method.

[0047] FIGS. 10A-10B present a representative use case of the novel methods disclosed herein using a response-relevant target map. FIG. 10A depicts a traditional clinical studywith no analytic plan for mapping neural drug response and no path to improve results of subsequent trials. In contrast, as shown in FIG. 10B, the methods disclosed herein may be used to facilitate the selection of subjects in a clinical trial that are more likely to respond to a therapeutic agent. In particular, such methods are employed to enrich for the precision selection of high responder candidate subjects.

[0048] FIG. 11 depicts a block diagram of a process for performing the computation framework relating to selecting subjects based on a response-relevant target map.

[0049] FIGS. 12A-12E depict representative analysis results of the novel methods disclosed herein for the purpose of increasing the effect size in a clinical trial. In particular, a response-relevant target map was used in this analysis, and the methods disclosed herein were applied to data from a trial investigating the effects of sertraline in depression (Trivedi, et al., J. Psychiatr. Res., 2016, 78:11-23). FIG. 12A shows the symptom weights obtained from fitting the symptom-to-alignment predictor function, here a linear regression model, to the trial data, and FIG. 12B provides a legend for the symptoms represented in FIG. 12A. FIG. 12C shows a histogram of the alignment scores; the 30% of subjects with the highest alignment scores (i.e., those subjects that would be selected for inclusion by the claimed method) are highlighted. FIG. 12D shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for inclusion by the claimed method. The black dashed line denotes the effect size when all subjects would be selected for inclusion (i.e., without application of the claimed method). FIG. 12E shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). As shown in FIG. 12E, a sample size reduction is enabled by the claimed methods.DETAILED DESCRIPTION

[0050] An illustrative embodiment of the present invention relates to the training and application of machine learning statistical models which select a candidate subject based on predicted response to a therapeutic agent. The model is first trained using a reference dataset with neural and behavioral data and a target neural map relating to the therapeutic agent, to determine a symptom-to-alignment predictor function that reflects the alignment between theneuro-behavioral data in the reference dataset and the neural target map. Specifically, a neural alignment score for each subject in the reference dataset is computed, using a quantitative measure of spatial similarity, such as Pearson’s correlation or Spearman’s rho, between the neural supervision target benchmark map and the individual’s neural feature map. Then, using data from all subjects in the reference dataset, a linear and / or non-linear machine learning statistical model is built to predict the subjects’ neural alignment score from their symptom measures. The trained model’s performance can be evaluated using methods such as independent sample replication or k-fold cross-validation.

[0051] The symptom-to-alignment predictor function from the trained model can then be applied to behavioral data from a new candidate subject in order to compute a predicted alignment score with the neural target map, which in turn reflects the subject’s probability of response to the therapeutic agent. Thus, subjects with a high predicted probability of response can be screened for the clinical trial to increase the probability of success.

[0052] FIGS. 1 through 12, wherein like parts are designated by like reference numerals throughout, illustrate an example embodiment or embodiments of the selection of individual subjects for likelihood of efficacy in clinical trials through machine learning informed supervision of neuroimaging and clinical scale variation, according to the present invention. Although the present invention will be described with reference to the example embodiment or embodiments illustrated in the figures, it should be understood that many alternative forms can embody the present inventions. One of skill in the art will additionally appreciate different ways to alter the parameters of the embodiment(s) disclosed in a manner still in keeping with the spirit and scope of the present invention.

[0053] With reference to FIG. 1, the invention comprises a statistical model (1) that is first trained on behavioral data (3) from a reference dataset (2) and the alignment scores (5) computed between neural data from the reference dataset (2) and a neural target map (4) relating to the experimental treatment of interest. The model (1) is then applied to behavioral data (3) from new candidate subjects (6) to compute predicted alignment scores with the neural target (7). An alignment score threshold value (8) is used to determine which candidate subjects are included or excluded from participation in the clinical trial based on their predicted alignment scores.

[0054] In operation, the invention disclosed herein can be used to train models using existing neuroimaging datasets, including existing data from prior clinical trials, and these models can subsequently be applied by clinical trial personnel to screen potential candidatesfor enrollment in new trials. As the trained model produces a symptom-to-alignment predictor function that can predict neural alignment scores from symptom / behavioral measures, only corresponding symptom / behavioral data from the relevant clinical scales, such as mood and depression symptoms from the Hamilton Depression Rating Scale (HAM- D); psychopathology symptoms from the Positive and Negative Syndrome Scale for Schizophrenia (PANSS); or cognitive performance from the Brief Assessment of Cognition in Schizophrenia (BACS) or Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog), need to be collected from new candidates.

[0055] Application of the present invention can increase the likelihood of a successful clinical trial by enriching the trial with subjects based on their predicted response to the experimental treatment and excluding subjects who are predicted to have a low likelihood of response to the experimental treatment and / or have a strong likelihood of response to placebo. Additionally, the models that are the subject of the present invention can be trained on existing data and applied to existing clinical trial workflows, requiring minimal disruption of clinical trial protocols. Furthermore, as the applied model only requires symptom / behavioral data to screen new candidates, and these symptom / behavioral measures may already be part of the data collection protocol for the clinical trial for efficacy and regulatory reasons, application of the present invention will require minimal to no additional data collection.

[0056] FIGS. 5A-5F present a representative use case applying the novel methods disclosed herein for the purpose of enriching a clinical trial with likely responders. As shown in FIGS. 5A-5F, these methods were applied to data from a trial investigating the effects of sertraline in depression (Trivedi, et al., J. Psychiatr. Res., 2016, 78:11-23). FIG. 5A shows the gene expression map of SLC29A4, a gene encoding a target for sertraline. In this specific example, gene expression across the whole brain including cerebral hemispheres and the subcortical regions were used, with the exception of the cerebellum due to incomplete data collection for this brain region in the EMB ARC dataset. The color scale of the gene expression map reflects the relative level of expression (Z-score) in each area of the brain, with yellow indicating a positive value, i.e., relatively high level of expression of the SLC29A4 gene, and bright blue areas indicating a negative value or relatively low level of expression. This gene expression map was used as the target map for the analyses presented in FIGS. 5A-5F. FIG. 5B shows the symptom weights obtained from fitting the symptom- to-alignment predictor function to the trial data. The symptom-alignment-predictor functionis a linear regression, represented by weights shown in FIG. 5B. FIG. 5C provides a legend for the symptoms represented in FIG. 5B. FIG. 5D shows a histogram of the alignment scores; the 20% of subjects with the highest alignment scores (i.e., those subjects that would be selected for inclusion in the clinical trial by the claimed method) are highlighted. The neural alignment score is a quantitative measure of spatial similarity, here a Pearson’s correlation, between the two neural maps. FIG. 5E shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for inclusion in the clinical trial by the claimed method. Effect sizes are calculated as the inverse-variance weighted average Cohen’s d obtained in the test-sets of a 10-times repeated 2-fold cross-validation of the model. Negative values represent a reduction in symptom score such that lower effect sizes are better. The black dashed line denotes the effect size when all subjects would be selected for inclusion (i.e., without application of the claimed method). FIG. 5F shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). The panel thus illustrates the sample size reduction enabled by the claimed method.

[0057] FIGS. 8A-8E present a representative use case of the novel methods disclosed herein for the purpose of excluding subjects with high predicted placebo response from a clinical trial. These methods disclosed herein were applied to data from a trial investigating the effects of sertraline in depression (Trivedi el al., J. Psychiatr. Res., 2016, 78:11-23). FIG. 8A shows the symptom weights obtained from fitting the symptom-to-alignment predictor function, here a linear regression model, to the trial data, and FIG. 8B provides a legend for the symptoms represented in FIG. 8A. FIG. 8C shows a histogram of the alignment scores; 80% of subjects with the lowest alignment scores are selected for exclusion by the claimed method, the remaining subjects, i.e., those that are selected for inclusion, are highlighted. FIG. 8D shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for exclusion by the claimed method. Effect sizes are calculated as the inverse-variance weighted average Cohen’s d obtained in the test-sets of a 10-times repeated 2-fold cross-validation of the model. Negative values represent a reduction in symptom score such that lower effect sizes are better. The black dashed line denotes the effect size when all subjects would be selectedfor inclusion (i.e., without application of the claimed method). FIG. 8E shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). FIG. 8E thus illustrates the required sample size reduction enabled by the claimed method.

[0058] FIGS. 9A-9F presents a representative use case of the novel methods disclosed herein for the purpose of reducing the required sample size in a clinical trial by both enriching it for likely responders and excluding subjects with high predicted placebo response. These methods disclosed herein were applied to data from a trial investigating the effects of sertraline in depression (Trivedi et al., J. Psychiatr. Res., 2016, 78: 11-23). In the current use case, an enrichment model (as in FIG. 5) and a placebo-filter model (as in FIG.8) were separately fitted to the data and subjects were selected for inclusion if both the enrichment model and placebo-filter model recommended a subject for inclusion. FIG. 9A shows the gene expression map of SLC29A4, a gene encoding a target for sertraline. This gene expression map was used as the target map for the enrichment model. FIG. 9B shows the symptom weights obtained from fitting the enrichment model, here a linear regression, to the trial data, and FIG. 9C shows the symptom weights obtained from fitting the placebo model, here a linear regression, to the trial data. FIG. 9D provides a legend for the symptoms represented in FIG. 9B and FIG. 9C. FIG. 9E shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for inclusion in the enrichment model and selected for exclusion in the placebo model. Effect sizes are calculated as the inverse- variance weighted average Cohen’ s d obtained in the test-sets of a 10-times repeated 2-fold cross-validation of the model. Negative values represent a reduction in symptom score such that lower effect sizes are better. The black dashed line denotes the effect size when all subjects would be selected for inclusion (i.e., without application of the claimed method). FIG. 9F shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). FIGS. 9A-9F thus illustrates the sample size reduction enabled by the claimed methods.

[0059] FIGS. 12A-12E present a representative use case of the novel methods disclosed herein relative to a traditional clinical trial. As shown in FIGS. 12A-12E, these methods were applied to data from a trial investigating the effects of sertraline in depression (Trivedi, et al., J. Psychiatr. Res., 2016, 78:11-23). FIG. 12A shows the symptom weights obtained from fitting the symptom-to-alignment predictor function, here a linear regression model, to the trial data, and FIG. 12B provides a legend for the symptoms represented in FIG. 12A. FIG. 12C shows a histogram of the alignment scores; the 30% of subjects with the highest alignment scores (i.e., those subjects that would be selected for inclusion by the claimed method) are highlighted. FIG. 12D shows the anticipated effect size as a function of the percentage of subjects with the highest alignment scores that would be selected for inclusion by the claimed method. Effect sizes are calculated as the inverse-variance weighted average Cohen’s d obtained in the test-sets of a 10-times repeated 2-fold cross-validation of the model. Negative values represent a reduction in symptom score such that lower effect sizes are better. The black dashed line denotes the effect size when all subjects would be selected for inclusion, i.e., without application of the claimed method. FIG. 12E shows the required sample size from a power calculation for a one-sided two-sample t-test assuming 80% power and a 5% false positive rate. The black dashed line denotes the required sample size when the effect size is determined without application of the claimed method (i.e., based on all available subjects independent of their symptom and biomarker profiles). FIG. 12E thus illustrates the sample size reduction enabled by the claimed method.DEFINITIONS

[0060] As used herein, the terms “drug” and “therapeutic agent” are used interchangeably to refer to any substance that is not food or part of a food and provides medical and / or health benefits, including prevention and / or treatment of disease.

[0061] As used herein, the term “neural supervision target benchmark map” means a representation of the brain where value(s) are assigned to specific locations / regions that reflect a molecular, circuit, mechanistic and / or biological property of the location; that is related to or believed to be related to the experimental treatment or therapeutic agent being studied; and that is used to guide the training of the machine learning statistical model.

[0062] As used herein, the term “neuro-behavior association dataset” means a dataset comprising one or more individuals with neural and symptom / behavioral data measures related to mental health or cognitive status.

[0063] As used herein, the term “neural target” means the receptor, circuit, or biological mechanism that is directly or indirectly affected by a therapeutic agent. Examples of neural targets include serotonin receptors, SLC29A / ENT transporter proteins, and presynaptic neurons.

[0064] As used herein, the term “neural alignment score” means a quantitative measure of spatial similarity, such as Pearson’s correlation or Spearman’s rho, between two neural maps, for example, between the neural supervision target benchmark map and a neural feature map(s) from the reference neuro-behavioral association dataset.

[0065] As used herein, the term “neural target data” means values measured, from one or more modalities in one or more individuals, relating to brain structure and / or function that is used to relate to the neural supervision target benchmark map during the training of the machine learning statistical model. Examples include, but are not limited to, functional connectivity derived from blood oxygen level dependent functional magnetic resonance imaging (BOLD MRI), and structural measures such as probabilistic tractography from diffusion weighted MRI.

[0066] As used herein, the term “neural target efficacy data” means neural target data, as defined above, that is collected from participants in a clinical trial and reflects the effect of the treatment (or placebo) underwent by said participants.

[0067] As used herein, the term “neural feature map” means a representation of the brain for a single subject where the value(s) in specific locations / regions reflect a measure(s) of a molecular, circuit, mechanistic and / or biological property of the location, or a score(s) derived from a statistical or mathematical process computed from such measures;

[0068] As used herein, the term “symptom measure” means a metric that quantifies the severity of an individual’s behavior and / or experience related to a psychiatric or neurological disorder or cognitive status. Examples include anhedonia, hallucinations, working memory deficits

[0069] As used herein, the term “symptom-to-alignment predictor function” means a linear or non-linear computation that encodes the relationship between symptom / behavior measures and neural alignment scores, and can therefore be used to predict one from theother. An exemplary linear computation is a linear regression, and exemplary non-linear computations include, but are not limited to, kernel regression and neural network.

[0070] It should be understood that embodiments of the present invention can be implemented in hardware, firmware, software, or a combination thereof. In such an embodiment, the various components and steps would be implemented in hardware, firmware, and / or software to perform the functions of the present invention. That is, the same piece of hardware, firmware, or module of software could perform one or more of the illustrated steps or components in the methods disclosed herein. The present invention can be implemented in one or more computer systems capable of carrying out the functionality described herein.100711 The computer systems disclosed herein may comprise or include one or more processors. Processors can be a special purpose or a general-purpose digital signal processor. In certain aspects, the processor is connected to a communication infrastructure, for example, a bus or network. Various software implementations are described in terms of this exemplary computer system. After reading this description, it will become apparent to a person skilled in the relevant art(s) how to implement the disclosure using other computer systems and / or computer architectures.

[0072] Computer system also includes a main memory, preferably random-access memory (RAM), and may also include a secondary memory. Secondary memory may include, for example, a hard disk drive and / or a removable storage drive, representing a floppy disk drive, a magnetic tape drive, an optical disk drive, solid-state disk, or the like. Removable storage drive reads from and / or writes to a removable storage unit in a well- known manner. Removable storage units may represent a floppy disk, magnetic tape, optical disk, solid-state disk, or the like, which is read by and written to by removable storage drive. As will be appreciated by persons skilled in the relevant art(s), removable storage units may include a computer usable storage medium having stored therein computer software and / or data.

[0073] In alternative implementations, secondary memory may include other similar means for allowing computer programs or other instructions to be loaded into a computer system. Such means may include, for example, a removable storage unit and an interface. Examples of such means may include a program cartridge and cartridge interface (such as that found in video game devices), a removable memory chip and associated socket, a thumbdrive and USB port, and other removable storage units and interfaces which allow software and data to be transferred from removable storage unit to the computer system.

[0074] The computer system may also include a communications interface. The communications interface allows software and data to be transferred between computer systems and external devices. Examples of communications interface may include a modem, a network interface (such as an Ethernet card), a communications port, a PCMCIA slot and card, etc. Software and data transferred via communications interface are in the form of signals which may be electronic, electromagnetic, optical, or other signals capable of being received by the communications interface. These signals are provided to the communications interface via a communications path. The communications path carries signals and may be implemented using wire or cable, fiber optics, a phone line, a cellular phone link, an RF link and other communications channels.

[0075] As used herein, the terms “computer program medium” and “computer readable medium” are used to generally refer to tangible storage media such as removable storage units or a hard disk installed in hard disk drive. These computer program products are means for providing software to computer system. Computer programs (also called computer control logic) are stored in main memory and / or secondary memory. Computer programs may also be received via communications interface. Such computer programs, when executed, enable the computer system to implement the present disclosed methods, as discussed herein. In particular, the computer programs, when executed, enable processor to implement the processes of the present inventions, such as any of the novel methods described herein. Accordingly, such computer programs represent controllers of the computer system. Where the disclosure is implemented using software, the software may be stored in a computer program product and loaded into computer system using removable storage drive, interface or communications interface.

[0076] In another embodiment, features of the disclosure are implemented primarily in hardware using, for example, hardware components such as application-specific integrated circuits (ASICs) and gate arrays. Implementation of a hardware state machine so as to perform the functions described herein will also be apparent to persons skilled in the relevant art(s).

[0077] Numerous modifications and alternative embodiments of the present invention will be apparent to those skilled in the art in view of the foregoing description. Accordingly, this description is to be construed as illustrative only and is for the purpose of teaching thoseskilled in the art the best mode for carrying out the present invention. Details of the structure may vary substantially without departing from the spirit of the present invention, and exclusive use of all modifications that come within the scope of the appended claims is reserved. Within this specification embodiments have been described in a way which enables a clear and concise specification to be written, but it is intended and will be appreciated that embodiments may be variously combined or separated without parting from the invention. It is intended that the present invention be limited only to the extent required by the appended claims and the applicable rules of law. The following specific examples are, therefore, to be construed as merely descriptive, and not limitative of the remainder of the disclosure in any way whatsoever.EXAMPLESExample 1: Using Symptom Measures to Enrich a Clinical Trial

[0078] FIGS. 5A-5F present a representative use case applying the novel method disclosed herein for the purpose of enriching a clinical trial for a specified neural supervision target benchmark map with likely responders. In this example, the method was applied to data from a trial investigating the effects of sertraline in depression (Trivedi, et al., J.Psychiatr. Res., 2016, 78:11-23) and the gene expression map of SLC29A4, a gene encoding a target for sertraline. This gene expression map was used as the neural supervision target benchmark map, and neural alignment scores were computed by calculating the similarity of each subject’s neural feature global brain connectivity (GBC) map to the SLC29A4 gene expression map. These neural alignment scores and the subjects’ symptom measures were then used to train a model that can predict a patient’s neural alignment to the SLC29A4 gene expression map based on symptoms alone. Here, the symptom-alignment-predictor function is a linear regression, represented by weights for each symptom measure. The symptom- alignment-predictor function was then applied to calculate predicted neural alignment scores from symptom measures, and a number of subjects with the highest alignment scores (e.g., top 20% or meeting a prespecified threshold) would be selected for inclusion in the clinical trial by the claimed method. The more selective the inclusion criteria (i.e., the higher the threshold for alignment scores of subjects that would be selected for inclusion in the clinical trial by the claimed method), the higher the anticipated effect size, and the lower the sample size required to power the clinical trial.Example 2: Excluding placebo responders from a clinical trial

[0079] FIGS. 8A-8E present a representative use case of the novel methods disclosed herein for the purpose of excluding subjects with high predicted placebo response from a clinical trial. The method disclosed herein was applied to data from a trial investigating the effects of sertraline in depression (Trivedi et al., J. Psychiatr. Res., 2016, 78:11-23). In this example, the neural supervision target benchmark map is representative of placebo response to serotonin, and the subjects with the lowest alignment scores are selected for exclusion by the claimed method (i.e., because they have a higher likelihood of being placebo responders). The remainder of the subjects would be included in the clinical trial.Example 3: Reducing the required sample size for powering a clinical trial by both enriching for likely responders and excluding placebo responders

[0080] FIGS. 9A-9F presents a representative use case of the novel methods disclosed herein for the purpose of reducing the required sample size in a clinical trial by both enriching it for likely responders and excluding subjects with high predicted placebo response. The methods disclosed herein were applied to data from a trial investigating the effects of sertraline in depression (Trivedi et al., J. Psychiatr. Res., 2016, 78:11-23). In this example, both an enrichment model (as in Example 1) and a placebo-filter model (as in Example 2) were separately fitted to the data and subjects were selected for inclusion if both the enrichment model and placebo-filter model recommended a subject for inclusion.Example 4: Using Results from a Previous Clinical Trial to Enrich a New Clinical Trial

[0081] FIGS. 12A-12E present a representative use case of the novel method disclosed herein relative to a traditional clinical trial. As shown in FIGS. 12A-12E, the method was applied to data from a trial investigating the effects of sertraline in depression (Trivedi, et al., J. Psychiatr. Res., 2016, 78:11-23). In this example, the neural supervision target benchmark map is a map of neural response to the experimental treatment from a previously conducted trial. By employing this method, a new clinical trial can leverage the information from the previously conducted trial to select patients that are enriched for high predicted neural alignment to the response map of the experimental treatment. This will increase the anticipated effect size and reduce the number of subjects required to power the clinical trial.

[0082] It is to be understood that the following claims are to cover all generic and specific features of the invention described herein, and all statements of the scope of the invention which, as a matter of language, might be said to fall therebetween.

Claims

CLAIMS1. A method of selecting a subject as a candidate subject based on their predicted response to an experimental treatment or therapeutic agent, the method comprising: a) retrieving, by at least one processor of a computing device, a neural supervision target benchmark map associated with the experimental treatment or therapeutic agent, i) wherein said neural supervision target benchmark map comprises neural target data collected from one or more individuals, the neural target data being associated with the experimental treatment or therapeutic agent, ii) wherein said neural supervision target benchmark map further comprises two-dimensional surfaces and subcortical volumetric structures in which the two-dimensional surfaces and the subcortical volumetric structures comprise one or more numerical values assigned to specific brain locations represented by the two-dimensional surfaces and the subcortical volumetric structures; b) retrieving, by the at least one processor of the computing device, a reference neuro-behavioral association dataset, i) wherein said reference neuro-behavioral association dataset comprises neural target data collected from one or more individuals, the neural target data including neural features and symptom features related to mental health or cognitive status; c) determining, by the at least one processor of the computing device, a symptom-to-alignment predictor function reflecting a relationship for the alignment between the neural supervision target benchmark map and the reference neuro- behavioral association dataset, i) wherein said determining comprises computing, via the at least one processor of the computing device, neural alignment scores for said reference neuro-behavioral association dataset relative to said neural supervision target benchmark map, ii) wherein a high absolute neural alignment score of said reference neuro-behavioral association dataset with said neural supervision target benchmark map is indicative of a statistical correspondence between the neuralfeatures in said reference neuro-behavioral dataset and the neural target data in said neural supervision target benchmark map, iii) wherein said neural alignment scores are used to compute the symptom-to-alignment predictor function that is indicative of a statistical correspondence between symptom features in the reference neuro-behavioral dataset and said neural alignment scores; and d) determining, by the at least one processor of the computing device, a neural alignment score of said subject, i) wherein said determining comprises assessing the symptom-to- alignment predictor function relative to said subject’s symptoms and / or assessing the neural supervision target benchmark map relative to said subject’s neural feature map; ii) wherein a high neural alignment score is indicative of a high probability of alignment with said neural supervision target benchmark map; and iii) wherein based on said high probability of alignment, said subject is given a quantitative likelihood of response to said experimental treatment or therapeutic agent.

2. The method of claim 1, wherein a high quantitative likelihood of response is indicative of the subject being selected as a candidate subject predicted to respond to said experimental treatment or therapeutic agent.

3. The method of claim 1, wherein a low quantitative likelihood of response is indicative of the subject being selected as a candidate subject that is not predicted to respond to said experimental treatment or therapeutic agent.

4. The method of claim 2, further comprising a step of selecting said candidate subject as likely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial.

5. The method of claim 3, further comprising a step of excluding said candidate subject as unlikely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial.

6. The method of claim 2, further comprising a step of administering said experimental treatment or therapeutic agent to said candidate subject selected as a candidate subject predicted to respond to said experimental treatment or therapeutic agent.

7. The method of claim 3, wherein if said subject is selected as a candidate subject not predicted to respond to said experimental treatment or therapeutic agent, said experimental treatment or therapeutic agent is not administered to said subject.

8. The method of claim 1, wherein said neural supervision target benchmark map comprises a brain-wide PET map of receptor occupancy for said therapeutic agent.

9. The method of claim 1, wherein said neural supervision target benchmark map comprises a pharmacological map associated with one or more receptor targets.

10. The method of claim 1, wherein said neural supervision target benchmark map comprises a gene expression map associated with one or more gene expression targets.

11. The method of claim 1, wherein said neural supervision target benchmark map comprises a previously computed neuro-behavioral variation map associated with one or more symptoms and / or indications.

12. The method of claim 1, wherein said neural supervision target benchmark map comprises a task-evoked neural map associated with one or more functions.

13. The method of claim 1, wherein the neural supervision target benchmark map is in a space where the subject’s left and right brain hemispheres are represented as surfaces and the subcortex is represented as a volume.

14. The method of claim 1, wherein the neural supervision target benchmark map and the reference neuro-behavioral association dataset are in the same space.

15. The method of claim 1, wherein the neural supervision target benchmark map and the reference neuro-behavioral association data set are parcellated according to an atlas of functionally-defined regions.

16. The method of claim 1, wherein the neural supervision target benchmark map is at the level of networks, areas, or individual vertices or voxels.

17. The method of claim 1, wherein the symptom-to-alignment predictor function is a linear regression model that produces a set of linear weightings that predict neural alignment scores from symptom measures.

18. The method of claim 1, wherein the symptom-to-alignment predictor function is a non-linear mapping that predicts neural alignment scores from symptom measures.

19. A method of selecting a subject predicted to respond to a placebo, the method comprising: a) retrieving, by at least one processor of a computing device, a placebo-relevant neural supervision target benchmark map, i) wherein said placebo-relevant neural supervision target benchmark map comprises neural target data collected from one or more individuals that participated in a clinical trial comprising a placebo arm; ii) wherein said placebo-relevant neural supervision target benchmark map further comprises two-dimensional surfaces and subcortical volumetric structures in which the two-dimensional surfaces and the subcortical volumetric structures comprise one or more numerical values assigned to specific brain locations represented by the two-dimensional surfaces and the subcortical volumetric structures; b) retrieving, by the at least one processor of the computing device, a reference neuro-behavioral association dataset, i) wherein said reference neuro-behavioral association dataset comprises neural target efficacy data collected from the one or more individuals that participated in the clinical trial and were randomized to the placebo arm, the neural target efficacy data including neural features and symptom features;c) determining, by the at least one processor of the computing device, a symptom-to-alignment predictor function reflecting a relationship for the alignment between the placebo-relevant neural supervision target benchmark map and the reference neuro-behavioral association dataset, i) wherein said determining comprises computing, via the at least one processor of the computing device, neural alignment scores for said reference neuro-behavioral association dataset relative to said placebo-relevant neural supervision target benchmark map, ii) wherein a high absolute neural alignment score of said reference neuro-behavioral association dataset with said placebo-relevant neural supervision target benchmark map is indicative of a statistical correspondence between the neural target efficacy data in the reference neuro-behavioral association dataset and the neural target data in the placebo-relevant neural supervision target benchmark map, iii) wherein said neural alignment scores are used to compute the symptom-to-alignment predictor function that is indicative of a statistical correspondence between symptom features in the reference neuro-behavioral dataset and said neural alignment scores; and d) determining, by the at least one processor of the computing device, a neural alignment score of said subject, i) wherein said determining comprises assessing the symptom-to- alignment predictor function relative to said subject’s symptoms and / or assessing said placebo-relevant neural supervision target benchmark map relative to said subject’s neural data, ii) wherein a high neural alignment score is indicative of a high probability of alignment with said placebo-relevant neural supervision target benchmark map, and iii) wherein based on said high probability of alignment, said subject is given a quantitative likelihood of response to placebo.

20. The method of claim 19, wherein the subject having a high neural alignment score has a high likelihood of being a placebo responder.

21. The method of claim 19, wherein the subject having a low neural alignment score has a low likelihood of being a placebo responder.

22. The method of claim 20, further comprising a step of excluding said subject as a likely placebo responder prior to randomization in a clinical trial.

23. The method of claim 19, further comprising selecting the subject as a candidate subject according to the method of claim 1.

24. The method of claim 19, wherein said placebo-relevant neural supervision target benchmark map comprises a brain-wide PET map of receptor occupancy associated with one or more receptor targets.

25. The method of claim 19, wherein said placebo-relevant neural supervision target benchmark map comprises a pharmacological map associated with one or more receptor targets.

26. The method of claim 19, wherein said placebo-relevant neural supervision target benchmark map comprises a gene expression map associated with one or more gene expression targets.

27. The method of claim 19, wherein said placebo-relevant neural supervision target benchmark map comprises a previously computed neuro-behavioral variation map associated with one or more symptoms and / or indications.

28. The method of claim 19, wherein said placebo-relevant neural supervision target benchmark map comprises a task-evoked neural map associated with one or more functions.

29. The method of claim 19, wherein the placebo-relevant neural supervision target benchmark map is in a space where the subject’s left and right brain hemispheres are represented as surfaces and the subcortex is represented as a volume.

30. The method of claim 19, wherein the placebo-relevant neural supervision target benchmark map and the reference neurobehavioral association dataset are in the same space.

31. The method of claim 19, wherein the placebo-relevant neural supervision target benchmark map and the reference neurobehavioral association dataset are parcellated according to an atlas of functionally-defined regions.

32. The method of claim 19, wherein the placebo-relevant neural supervision target benchmark map is at the level of networks, areas, or individual vertices or voxels.

33. The method of claim 19, wherein the symptom-to- alignment predictor function is a linear regression model that produces a set of linear weightings that predict neural alignment scores from symptom measures.

34. The method of claim 19, wherein the symptom-to-alignment predictor function is a non-linear mapping that predicts neural alignment scores from symptom measures.

35. A method of selecting a subject as a candidate subject based on their predicted response to an experimental treatment or a therapeutic agent, the method comprising: a) retrieving, by at least one processor of a computing device, a neural supervision target benchmark map, i) wherein said neural supervision target benchmark map comprises neural target efficacy data collected from one or more individuals that participated in a clinical trial of the experimental treatment or therapeutic agent, the clinical trial comprising a treatment arm and a placebo arm; ii) wherein said neural supervision target benchmark map further comprises two-dimensional surfaces and subcortical volumetric structures in which the two-dimensional surfaces and the subcortical volumetric structures comprise one or more numerical values assigned to specific brain locations represented by the two-dimensional surfaces and the subcortical volumetric structures;b) retrieving, by the at least one processor of the computing device, a reference neuro-behavioral association dataset, i) wherein said reference neuro-behavioral association dataset comprises baseline clinical data collected from one or more individuals, the baseline clinical data including neural target efficacy data and symptom features related to mental health or cognitive status; c) determining, by the at least one processor of the computing device, a symptom-to-alignment predictor functions reflecting a relationship for the alignment between the neural supervision target benchmark map and the reference neuro- behavioral association dataset, i) wherein said determining comprises computing, via the at least one processor of the computing device, neural alignment scores for said reference neuro-behavioral association dataset relative to said neural supervision target benchmark map, ii) wherein a high absolute neural alignment score of said reference neuro-behavioral association dataset with said neural supervision target benchmark map is indicative of a statistical correspondence between the neural target efficacy data in the reference neuro-behavioral association dataset and the neural target efficacy data in the neural supervision target benchmark map, iii) wherein said neural alignment scores are used to compute the symptom-to-alignment predictor function that is indicative of a statistical correspondence between symptom features in the reference neuro-behavioral dataset and said neural alignment scores; and d) determining, by the at least one processor of the computing device, a neural alignment score of said subject, i) wherein said determining comprises assessing the symptom-to- alignment predictor function relative to said subject’s symptoms and / or assessing the neural supervision target benchmark map relative to said subject’s neural data; ii) wherein a high neural alignment score is indicative of a high probability of alignment with said neural supervision target benchmark map, and iii) wherein based on said high probability of alignment, said subject is given a quantitative likelihood of response to said experimental treatment or therapeutic agent.

36. The method of claim 35, wherein a high quantitative likelihood of response is indicative of the subject being selected as a candidate subject predicted to respond to said experimental treatment or therapeutic agent.

37. The method of claim 35, wherein a low quantitative likelihood of response is indicative of the subject being selected as a candidate subject that is not predicted to respond to said experimental treatment or therapeutic agent.

38. The method of claim 36, further comprising a step of selecting said candidate subject as likely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial.

39. The method of claim 37, further comprising a step of excluding said candidate subject as unlikely to respond to said experimental treatment or therapeutic agent prior to randomization in a clinical trial.

40. The method of claim 36, further comprising a step of administering said experimental treatment or therapeutic agent to said candidate subject.

41. The method of claim 37, wherein if said subject is selected as a candidate subject that is not predicted to respond to said experimental treatment or therapeutic agent, said experimental treatment or therapeutic agent is not administered to said subject.

42. The method of claim 35, further comprising selecting the subject predicted to respond to a placebo according to the method of claim 19.

43. The method of claim 35, wherein said neural supervision target benchmark map comprises a brain-wide PET map of receptor occupancy for said therapeutic agent.

44. The method of claim 35, wherein said neural supervision target benchmark map comprises a pharmacological map associated with one or more receptor targets.

45. The method of claim 35, wherein said neural supervision target benchmark map comprises a gene expression map associated with one or more gene expression targets.

46. The method of claim 35, wherein said neural supervision target benchmark map comprises a previously computed neuro-behavioral variation map associated with one or more symptoms and / or indications.

47. The method of claim 35, wherein said neural supervision target benchmark map comprises a task-evoked neural map associated with one or more functions.

48. The method of claim 35, wherein the neural supervision target benchmark map is in a space where the subject’s left and right brain hemispheres are represented as surfaces and the subcortex is represented as a volume.

49. The method of claim 35, wherein the neural supervision target benchmark map and the reference neurobehavioral association dataset are in the same space.

50. The method of claim 35, wherein the neural supervision target benchmark map and the reference neurobehavioral association dataset are parcellated according to an atlas of functionally-defined regions51. The method of claim 35, wherein the neural supervision target benchmark map is at the level of networks, areas, or individual vertices or voxels.

52. The method of claim 35, wherein the symptom-to- alignment predictor function is a linear regression model that produces a set of linear weightings that predict neural alignment scores from symptom measures.

53. The method of claim 35, wherein the symptom-to-alignment predictor function is a non-linear mapping that predicts neural alignment scores from symptom measures.

Images