Systems and methods for predicting drug-induced liver injury

Abstract

Assays for determining a risk that a compound will induce a liver injury in a subject are provided. A dataset is obtained that comprises a plurality of cellular constituent abundance values for a plurality of cellular constituents for each cell in a plurality of cells exposed to the test chemical compound for a first period of time. The dataset is used to determine a pathway value for each module in a plurality of pathway modules. Each module comprises an independent subset of the plurality of cellular constituents. The pathway values are inputted into one or more first models thereby retrieving one or more first values. The abundance values are inputted into one or more second models thereby retrieving one or more second values. The one or more first values and one or more second values determine a risk that the compound will induce the liver injury.

Description

SYSTEMS AND METHODS FOR PREDICTING DRUG-INDUCED LIVER INJURYCROSS-REFERENCE TO RELATED PATENT APPLICATIONS

[0001] The present application claims priority to United States Provisional Patent Application No. 63 / 515,536 entitled “Systems and Methods for Predicting Drug-Induced Liver Injury,” filed July 25, 2023, which is hereby incorporated by reference.TECHNICAL FIELD

[0002] The present invention relates generally to systems and methods for identifying compounds that are likely to be associated with drug-induced liver injury.BACKGROUND

[0003] Drug-induced liver injury (DILI) may cause acute liver failure and is a common reason for clinical attrition of drug candidates, precautionary warnings, and post-market withdrawal of approved drugs. See, Kaplowitz, 2004, “Drug-induced liver injury,” Clin. Infect. Dis. 38, S44-S48, doi: 10.1086 / 381445; and Senior, 2007, “Drug hepatotoxicity from a regulatory perspective,” Clin. Liver Dis. 11, pp. 507-524, doi:10.1016 / j.cld.2007.06.002, each of which is hereby incorporated by reference. Early assessment of DILI risk is an important aspect of drug development, but doing so is challenging prior to clinical trials due to many complex factors that give rise to liver damage, further complicated by metabolic differences between animals and humans. In practice, attributing predictive risk of DILI to any given compound during drug discovery is considered arbitrary and difficult. See, for example, Hoofnagle and Bjbrnsson, 2019, “Drug-induced liver injury - Types and phenotypes,” N. Engl. J. Med. 381, 264-273. doi: 10.1056 / nejmral816149e, which is hereby incorporated by reference.

[0004] In view of the significance of DILI to drug development and difficulties in attributing DILI-related risk to drug candidates, what is needed in the art are improved systems, methods and assays for more accurately assessing the degree of risk that an investigational compound for drug discovery may induce liver injury in a human subject.SUMMARY

[0005] The present disclosure addresses above-identified shortcomings. A method is provided of assessing risk that a compound for drug discovery may induce liver injury in a human subject.

[0006] In some embodiments, the test chemical compound is an organic compound, particularly a small-molecule organic compound, having a molecular weight of less than 2000 Daltons. In some embodiments, the test chemical compound is an organic compound having a molecular weight of less than 1000 Daltons. In some embodiments, the test chemical compound is an organic compound that satisfies each of the Lipinski rule of five criteria. In some embodiments, the test chemical compound is an organic compound that satisfies at least three criteria of the Lipinski rule of five criteria. In some embodiments, the test chemical compound is a polypeptide, protein, oligonucleotide, RNA-based therapeutic, antibody, or antibody-drug conjugate. In some embodiments, the liver injury is hepatitis, hepatotoxicityjaundice, hepatic fibrosis, cirrhosis, hepatomegaly, alcoholic and nonalcoholic liver fatty liver disease and / or choluria. In some embodiments, the liver injury is characterized by an elevation of one or more liver enzymes, cholestasis, and / or alcoholic stools.

[0007] A cellular constituent dataset is obtained in electronic form. The dataset comprises a plurality of cellular constituent abundance values for a plurality of cellular constituents in a first plurality of cells that have been exposed to the test chemical compound for a first period of time.

[0008] In some embodiments, each cellular constituent in the plurality of cellular constituents is a particular gene, a particular mRNA associated with a gene, a carbohydrate, a lipid, an epigenetic feature, a metabolite, a protein, or a combination thereof.

[0009] In some embodiments, the first plurality of cells is primary human hepatocytes. In some embodiments, the plurality of cellular constituent abundance values is determined by single-cell ribonucleic acid (RNA) sequencing (scRNA-seq) data of the first plurality of cells. In some embodiments, the plurality of cellular constituent abundance values is determined by bulk RNA sequencing. In some embodiments, the plurality of cellular constituent abundance values is determined by single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq), CyTOF / SCoP, E-MS / Abseq, miRNA-seq, CITE-seq, or any combination thereof. In some embodiments, the plurality of cellular constituents consists of between 100 and 35,000 cellular constituents.

[0010] The cellular constituent dataset is used to determine a respective pathway value for each pathway module in a plurality of pathway modules, thereby obtaining a plurality of pathway values. Each respective pathway module in the plurality of pathway modules comprises an independent subset of the plurality of cellular constituents. Each independent subset of the plurality of cellular constituents for each respective pathway module comprises two or more cellular constituents. Each cellular constituent in the plurality of cellular constituents is in at least one independent subset of the plurality of cellular constituents associated with a pathway module in the plurality of pathway modules. The plurality of cellular constituents comprises 10 or more cellular constituents. In some embodiments, the plurality of cellular constituent abundance values comprises 5000 or more cellular constituent abundance values. In some embodiments, the plurality of cellular constituent abundance values comprises 5000, 10,000, 15,000, 20,000, 25,000, or 30,000 or more cellular constituent abundance values. In some embodiments each cellular constituent is a different gene.

[0011] In some embodiments, each independent subset of the plurality of cellular constituents for each respective pathway module comprises between two cellular constituents and three hundred cellular constituents. In some embodiments, each respective pathway value of each respective pathway module is a measure of central tendency of an abundance of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents of the respective pathway, across the cellular constituent dataset. In some embodiments, each respective pathway value of each respective pathway module is a negative base 10 logarithm of a false discovery rate determined based on a distribution of abundances of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents of the respective pathway module, across the cellular constituent dataset, relative to abundances of cellular constituents in the cellular constituent dataset.

[0012] In some embodiments, the method further comprises identifying the plurality of pathway modules by a process comprising: obtaining one or more first datasets in electronic form, the one or more first datasets comprising: for each respective primary human hepatocyte cell in a plurality of primary human hepatocyte cells, wherein the plurality of primary human hepatocyte cells comprises twenty or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states: for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell, identifyingthe plurality of cellular constituents, in the one or more first datasets, as a plurality of differentially expressed cellular constituents across the one or more first datasets; and associating each independent subset of cellular constituents in the plurality of cellular constituents with a corresponding pathway for a pathway module in the plurality of pathway modules.

[0013] In some embodiments, an annotated cell state in the plurality of annotated cell states is an exposure of a cell in the plurality of primary human hepatocyte cells to a training compound under an exposure condition (e.g., the exposure condition is a duration of exposure, a concentration of a training compound, or a combination of a duration of exposure and a concentration of the training compound).

[0014] In some embodiments, the plurality of annotated cell states includes exposure to a first training compound at a plurality of different concentrations (e.g., the plurality of different concentrations range from (i) a Cmax or an imputed Cmax of the first training compound to (ii) an ICio of the first training compound).

[0015] In some embodiments, the plurality of annotated cell states includes exposure to a first training compound at a plurality of different durations of exposure. In some embodiments, the plurality of annotated cell states includes exposure to a first training compound each for the same duration of exposure. In some such embodiments an optimal duration is determined through experimentation.

[0016] In some embodiments, the plurality of annotated cell states includes exposure to each respective training compound in a plurality of training compounds at a corresponding plurality of different concentrations (e.g., the corresponding plurality of different concentrations range from (i) a Cmax of the respective training compound to (ii) an ICio of the respective training compound).

[0017] In some embodiments, the plurality of annotated cell states includes exposure to each training compound in the plurality of training compounds at a plurality of different durations of exposure.

[0018] In some embodiments, the plurality of pathway modules consists of between 10 and 2000 pathway modules.

[0019] The plurality of pathway values is inputted into one or more first models thereby retrieving, as output from the one or more first models, one or more first value.

[0020] In some embodiments, each first model in the one or more first models is a logistic regression model. In some embodiments, each first model in the one or more firstmodel is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model.

[0021] In some embodiments, the one or more first models comprises 10 or more first modules, 20 or more first modules, 30 or more first modules, 40 or more first modules, 50 or more first modules, 60 or more first modules, 70 or more first modules, 80 or more first modules, or 90 or more first modules.

[0022] The plurality of cellular constituent abundance values is inputted into one or more second models thereby retrieving, as output from the one or more second models, one or more second values.

[0023] In some embodiments, the one or more second models is a single logistic regression model. In some embodiments, the one or more second models a plurality of logistic regression models. In some such embodiments, a first subset of logistic regression models in the plurality of second models is a plurality of LI regression models, and a second subset of logistic regression models in the plurality of second models is a plurality of L2 regression models. In some embodiments, the plurality of second models is between two and 1000 logistic regression models. In some embodiments, the one or more second models comprises 10 or more second modules, 20 or more second modules, 30 or more second modules, 40 or more second modules, 50 or more second modules, 60 or more second modules, 70 or more second modules, 80 or more second modules, or 90 or more second modules.

[0024] In some embodiments, each second model in the one or more second models is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model.

[0025] The first value and the one or more second values are used to determine a risk that the test chemical compound will induce the liver injury in the human subject.

[0026] In some embodiments, the risk is in the form of a probability that the test chemical compound will induce the liver injury in the human subject.

[0027] In some embodiments, the risk is one level of risk from among an enumerated set of levels of risk that the test chemical compound will induce the liver injury in the humansubject. In some such embodiments, the enumerated set of levels of risk consists of a low, a medium, and a high level of risk that the test chemical compound will induce the liver injury in the human subject.

[0028] In some embodiments, the risk in the form of the lowest concentration of the test chemical compound predicted by the one or more first models and the one or more second models to induce the liver injury in the human subject.

[0029] In some embodiments, the method further comprises determining a safety margin of the test chemical compound using the lowest concentration of the test chemical compound predicted by the one or more first models and the one or more second models to induce the liver injury in the human subject (e.g., determining the safety margin comprises dividing the risk by a Cmax of the test chemical compound).

[0030] In some embodiments, the method informs the selection of one or more human subjects for treatment with the test chemical compound and / or selection of one or more human subjects for continuation or discontinuation of treatment with the test chemical compound.

[0031] In some embodiments, the method informs the dosing amount, duration, and / or frequency of the test chemical compound in one or more human subjects for treatment.

[0032] In some embodiments, the method informs the design of a clinical trial, the clinical trial comprising the use of the test chemical compound.

[0033] In some embodiments, the method informs the type and / or level of acceptable adverse events, one or more inclusion criteria, and / or one or more exclusion criteria for the clinical trial.

[0034] In some embodiments, the method informs one or more modifications of the clinical trial while the clinical trial is in progress.

[0035] In some embodiments, the method informs the design of an adaptive clinical trial, the adaptive clinical trial comprising the use of the test chemical compound.

[0036] In some embodiments, the method informs the type and / or level of acceptable adverse events, one or more inclusion criteria, and / or one or more exclusion criteria for the adaptive clinical trial.

[0037] In some embodiments, the method informs one or more modifications of the adaptive clinical trial while the clinical trial is in progress.

[0038] In some embodiments, the method informs classification of the test chemical compound as a compound suitable for human therapy and / or testing or a compound unsuitable human therapy and / or testing.

[0039] In some embodiments, the method further comprises formulating the test chemical compound for use in a therapy.

[0040] Another aspect of the present disclosure provides a computer system having one or more processors, and memory storing one or more programs for execution by the one or more processors, the one or more programs comprising instructions for performing any of the methods and / or embodiments disclosed herein.

[0041] Another aspect of the present disclosure provides a non-transitory computer readable storage medium storing one or more programs configured for execution by a computer, the one or more programs comprising instructions for carrying out any of the methods and / or embodiments disclosed herein.

[0042] Additional aspects and advantages of the present disclosure will become readily apparent to those skilled in this art from the following detailed description, wherein only illustrative embodiments of the present disclosure are shown and described. As will be realized, the present disclosure is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the disclosure. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The embodiments disclosed herein are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings. Like reference numerals refer to corresponding parts throughout the drawings.

[0044] Figures 1 A and IB illustrate a block diagram of an exemplary system and computing device, in accordance with an embodiment of the present disclosure.

[0045] Figures 2A, 2B, 2C, 2D, 2E, 2F, 2G, and 2H collectively provide a flow chart of processes and features of an example method for assessing a risk that a test chemical compound will induce a liver injury in a human subject, in accordance with various embodiments of the present disclosure.

[0046] Figure 3 illustrates how transcriptomics from primary human hepatocytes can be used to predict DILI safety margins to guide hit selection and lead optimization in a drug discovery program in accordance with an embodiment of the present disclosure.

[0047] Figure 4 illustrates the respective pathway values, in the form of a color coded heatmap, for a plurality of pathway modules formed from training data in accordance with an embodiment of the present disclosure.

[0048] Figure 5 shows the performance of a trained ensemble model at predicting DILI toxicity in accordance with an embodiment of the present disclosure.

[0049] Figure 6 illustrates the predicted safety margin for Amiodarone that was calculated using a trained ensemble model in accordance with an embodiment of the present disclosure.

[0050] Figure 7 illustrates the expression gene signatures measured for select test compounds at each of four different exposure concentrations, in accordance with an embodiment of the present disclosure.

[0051] Figures 8A and 8B illustrate the pathway values for pathway modules for select test compounds at each of four different exposure concentrations, in accordance with an embodiment of the present disclosure.

[0052] Figure 9 illustrates how Erthromycen and Labelatol fails to cause expression of numerous DILI markers.

[0053] Figure 10 plots the predicted safety margin for toxicity determined by an ensemble model of the present disclosure (Y-axis) against the true DILI toxicity (X-axis) for 30 test compounds, in accordance with an embodiment of the present disclosure.

[0054] Figure 11 plots the predicted safety margin for toxicity determined by the ensemble model (X-axis) against the known clinical therapeutic index (Y-axis) for the 30 test compounds, in accordance with an embodiment of the present disclosure.DETAILED DESCRIPTION

[0055] Introduction.

[0056] Given the above background, the present disclosure provides assays for determining a risk that a compound will induce a liver injury in a subject. A plurality of cellular constituent abundance values for a plurality of cellular constituents in a plurality of cells that have been exposed to the test chemical compound for a first period of time is obtained. This data is used to determine a pathway value for each module in a plurality ofpathway modules. Each such module comprises an independent subset of the plurality of cellular constituents. The pathway values are inputted into one or more first models thereby retrieving one or more first values. The plurality of cellular constituent abundance values is inputted into one or more second models thereby retrieving one or more second values. The one or more first values and one or more second values are used to determine a risk that the compound will induce the liver injury.

[0057] Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent to one of ordinary skill in the art that the present disclosure may be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments.

[0058] Plural instances may be provided for components, operations or structures described herein as a single instance. Finally, boundaries between various components, operations, and data stores are somewhat arbitrary, and particular operations are illustrated in the context of specific illustrative configurations. Other forms of functionality are envisioned and may fall within the scope of the implementation(s). In general, structures and functionality presented as separate components in the example configurations may be implemented as a combined structure or component. Similarly, structures and functionality presented as a single component may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of the implementation(s).

[0059] It will also be understood that, although the terms “first,” “second,” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first dataset could be termed a second dataset, and, similarly, a second dataset could be termed a first dataset, without departing from the scope of the present invention. The first dataset and the second dataset are both datasets, but they are not the same dataset.

[0060] The terminology used herein is for the purpose of describing particular implementations only and is not intended to be limiting of the claims. As used in the description of the implementations and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term “and / or” as used herein refers to andencompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “comprises” and / or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0061] As used herein, the term “if’ may be construed to mean “when” or “upon” or “in response to determining” or “in accordance with a determination” or “in response to detecting,” that a stated condition precedent is true, depending on the context. Similarly, the phrase “if it is determined (that a stated condition precedent is true)” or “if (a stated condition precedent is true)” or “when (a stated condition precedent is true)” may be construed to mean “upon determining” or “in response to determining” or “in accordance with a determination” or “upon detecting” or “in response to detecting” that the stated condition precedent is true, depending on the context.

[0062] Furthermore, when a reference number is given an “zth” denotation, the reference number refers to a generic component, set, or embodiment. For instance, a cellular- component termed “cellular-component z” refers to the zthcellular-component in a plurality of cellular-components.

[0063] In the interest of clarity, not all of the routine features of the implementations described herein are shown and described. It will be appreciated that, in the development of any such actual implementation, numerous implementation-specific decisions are made in order to achieve the designer’s specific goals, such as compliance with use case- and business-related constraints, and that these specific goals will vary from one implementation to another and from one designer to another. Moreover, it will be appreciated that such a design effort might be complex and time-consuming, but nevertheless be a routine undertaking of engineering for those of ordering skill in the art having the benefit of the present disclosure.

[0064] Some portions of this description describe the embodiments of the invention in terms of algorithms and symbolic representations of operations on information. These algorithmic descriptions and representations are commonly used by those skilled in the data processing arts to convey the substance of their work effectively to others skilled in the art. These operations, while described functionally, computationally, or logically, are understood to be implemented by computer programs or equivalent electrical circuits, microcode, or the like.

[0065] The language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the invention be limited not by this detailed description, but rather by any claims that issue on an application based hereon. Accordingly, the disclosure of the embodiments of the invention is intended to be illustrative, but not limiting, of the scope of the invention.

[0066] In general, terms used in the claims and the specification are intended to be construed as having the plain meaning understood by a person of ordinary skill in the art. Certain terms are defined below to provide additional clarity. In case of conflict between the plain meaning and the provided definitions, the provided definitions are to be used.

[0067] Any terms not directly defined herein shall be understood to have the meanings commonly associated with them as understood within the art of the invention. Certain terms are discussed herein to provide additional guidance to the practitioner in describing the compositions, devices, methods and the like of aspects of the invention, and how to make or use them. It will be appreciated that the same thing may be said in more than one way. Consequently, alternative language and synonyms may be used for any one or more of the terms discussed herein. No significance is to be placed upon whether or not a term is elaborated or discussed herein. Some synonyms or substitutable methods, materials and the like are provided. Recital of one or a few synonyms or equivalents does not exclude use of other synonyms or equivalents, unless it is explicitly stated. Use of examples, including examples of terms, is for illustrative purposes only and does not limit the scope and meaning of the aspects of the invention herein.

[0068] Definitions.

[0069] As used herein, the term “about” or “approximately” means within an acceptable error range for the particular value as determined by one of ordinary skill in the art, which depends in part on how the value is measured or determined, e.g., the limitations of the measurement system. For example, in some embodiments “about” means within 1 or more than 1 standard deviation, per the practice in the art. In some embodiments, “about” means a range of ±20%, ±10%, ±5%, or ±1% of a given value. In some embodiments, the term “about” or “approximately” means within an order of magnitude, within 5-fold, or within 2- fold, of a value. Where particular values are described in the application and claims, unless otherwise stated the term “about” meaning within an acceptable error range for the particular value can be assumed. All numerical values within the detailed description herein are modified by “about” the indicated value, and consider experimental error and variations thatwould be expected by a person having ordinary skill in the art. The term “about” can have the meaning as commonly understood by one of ordinary skill in the art. In some embodiments, the term “about” refers to ±10%. In some embodiments, the term “about” refers to ±5%.

[0070] As used herein, the terms “abundance,” “abundance level,” or “expression level” refers to an amount of a cellular constituent (e.g., a gene product such as an RNA species, e.g., mRNA or miRNA, or a protein molecule) present in one or more cells, or an average amount of a cellular constituent present across multiple cells. When referring to mRNA or protein expression, the term generally refers to the amount of any RNA or protein species corresponding to a particular genomic locus, e.g., a particular gene. However, in some embodiments, an abundance can refer to the amount of a particular isoform of an mRNA or protein corresponding to a particular gene that gives rise to multiple mRNA or protein isoforms. The genomic locus can be identified using a gene name, a chromosomal location, or any other genetic mapping metric.

[0071] As used interchangeably herein, a “cell state” or “biological state” refers to a state or phenotype of a cell or a population of cells. For example, a cell state can be healthy or diseased. A cell state can be one of a plurality of diseases. A cell state can be a response to a compound treatment and / or a differentiated cell lineage. A cell state can be characterized by a measure (e.g., an activation, expression, and / or measure of abundance) of one or more cellular constituents, including but not limited to one or more genes, one or more proteins, and / or one or more biological pathways.

[0072] As used herein, a “cell state transition” or “cellular transition” refers to a transition in a cell’s state from a first cell state to a second cell state. In some embodiments, the second cell state is an altered cell state (e.g., a healthy cell state to a diseased cell state). In some embodiments, one of the respective first cell state and second cell state is an unperturbed state and the other of the respective first cell state and second cell state is a perturbed state caused by an exposure of the cell to a condition. The perturbed state can be caused by exposure of the cell to a compound. A cell state transition can be marked by a change in cellular constituent abundance in the cell, and thus by the identity and quantity of cellular constituents (e.g., mRNA, transcription factors) produced by the cell (e.g., a perturbation signature).

[0073] As used herein, the term “dataset” in reference to cellular constituent abundance measurements for a cell or a plurality of cells can refer to a high-dimensional set of data collected from a single cell (e.g., a single-cell cellular constituent abundance dataset) in some contexts. In other contexts, the term “dataset” can refer to a plurality of high-dimensionalsets of data collected from single cells (e.g., a plurality of single-cell cellular constituent abundance datasets), each set of data of the plurality collected from one cell of a plurality of cells.

[0074] As used herein, the term “differential abundance” or “differential expression” refers to differences in the quantity and / or the frequency of a cellular constituent present in a first entity (e.g., a first cell, plurality of cells, and / or sample) as compared to a second entity (e.g., a second cell, plurality of cells, and / or sample). In some embodiments, a first entity is a sample characterized by a first cell state e.g., a diseased phenotype) and a second entity is a sample characterized by a second cell state e.g., a normal or healthy phenotype). For example, a cellular constituent can be a polynucleotide e.g., an mRNA transcript) which is present at an elevated level or at a decreased level in entities characterized by a first cell state compared to entities characterized by a second cell state. In some embodiments, a cellular constituent can be a polynucleotide which is detected at a higher frequency or at a lower frequency in entities characterized by a first cell state compared to entities characterized by a second cell state. A cellular constituent can be differentially abundant in terms of quantity, frequency or both. In some instances, a cellular constituent is differentially abundant between two entities if the amount of the cellular constituent in one entity is statistically significantly different from the amount of the cellular constituent in the other entity. For example, a cellular constituent is differentially abundant in two entities if it is present at least about 120%, at least about 130%, at least about 150%, at least about 180%, at least about 200%, at least about 300%, at least about 500%, at least about 700%, at least about 900%, or at least about 1000% greater in one entity than it is present in the other entity, or if it is detectable in one entity and not detectable in the other. In some instances, a cellular constituent is differentially expressed in two sets of entities if the frequency of detecting the cellular constituent in a first subset of entities (e.g., cells representing a first subset of annotated cell states) is statistically significantly higher or lower than in a second subset of entities (e.g., cells representing a second subset of annotated cell states). For example, a cellular constituent is differentially expressed in two sets of entities if it is detected at least about 120%, at least about 130%, at least about 150%, at least about 180%, at least about 200%, at least about 300%, at least about 500%, at least about 700%, at least about 900%, or at least about 1000% more frequently or less frequently observed in one set of entities than the other set of entities.

[0075] As used herein, the term “sample,” “biological sample,” or “patient sample,” refers to any sample taken from a subject, which can reflect a biological state associated withthe subject. Examples of samples include, but are not limited to, blood, whole blood, plasma, serum, urine, cerebrospinal fluid, fecal, saliva, sweat, tears, pleural fluid, pericardial fluid, or peritoneal fluid of the subject. A sample can include any tissue or material derived from a living or dead subject. A sample can be a cell-free sample. A sample can comprise one or more cellular constituents. For instance, a sample can comprise a nucleic acid (e.g., DNA or RNA) or a fragment thereof, or a protein. The term “nucleic acid” can refer to deoxyribonucleic acid (DNA), ribonucleic acid (RNA) or any hybrid or fragment thereof. The nucleic acid in the sample can be a cell-free nucleic acid. A sample can be a liquid sample or a solid sample (e.g., a cell or tissue sample). A sample can be a bodily fluid. A sample can be a stool sample. A sample can be treated to physically disrupt tissue or cell structure (e.g., centrifugation and / or cell lysis), thus releasing intracellular components into a solution which can further contain enzymes, buffers, salts, detergents, and the like which can be used to prepare the sample for analysis.

[0076] As used interchangeably herein, the term “classifier”, “model”, algorithm, “regressor”, and / ”or classifier” refers to a machine learning model or algorithm.

[0077] In some embodiments, a model used in the present disclosure is supervised machine learning model. Nonlimiting examples of supervised learning algorithms that can be used as a model in the present disclosure include, but are not limited to, logistic regression, neural networks, support vector machines, Naive Bayes algorithms, nearest neighbor algorithms, random forest algorithms, decision tree algorithms, boosted trees algorithms, multinomial logistic regression algorithms, linear models, linear regression, GradientBoosting, mixture models, hidden Markov models, Gaussian NB algorithms, linear discriminant analysis, or any combinations thereof. In some embodiments, a model is a multinomial classifier algorithm. In some embodiments, a model is a 2-stage stochastic gradient descent (SGD) model. In some embodiments, a model is a deep neural network (e.g, a deep-and-wide sample-level model). In some embodiments, a classifier or model of the present disclosure has 25 or more, 100 or more, 1000 or more 10,000 or more, 100,000 or more or 1 x 106or more parameters and thus, in some embodiments, the calculations of the model cannot be mentally performed.

[0078] Moreover, as used herein, the term “parameter” refers to any coefficient or, similarly, any value of an internal or external element (e.g., a weight and / or a hyperparameter) in an algorithm, model, regressor, and / or classifier that can affect (e.g., modify, tailor, and / or adjust) one or more inputs, outputs, and / or functions in the algorithm, model, regressor and / or classifier. For example, in some embodiments, a parameter refers toany coefficient, weight, and / or hyperparameter that can be used to control, modify, tailor, and / or adjust the behavior, learning, and / or performance of an algorithm, model, regressor, and / or classifier. In some instances, a parameter is used to increase or decrease the influence of an input (e.g., a feature) to an algorithm, model, regressor, and / or classifier. As a nonlimiting example, in some embodiments, a parameter is used to increase or decrease the influence of a node (e.g., of a neural network), where the node includes one or more activation functions. Assignment of parameters to specific inputs, outputs, and / or functions is not limited to any one paradigm for a given algorithm, model, regressor, and / or classifier but can be used in any suitable algorithm, model, regressor, and / or classifier architecture for a desired performance. In some embodiments, a parameter has a fixed value. In some embodiments, a value of a parameter is manually and / or automatically adjustable. In some embodiments, a value of a parameter is modified by a validation and / or training process for an algorithm, model, regressor, and / or classifier (e.g., by error minimization and / or backpropagation methods). In some embodiments, an algorithm, model, regressor, and / or classifier of the present disclosure includes a plurality of parameters. In some embodiments, the plurality of parameters is n parameters, where: n > 2; n > 5; n > 10; n > 25; n > 40; n > 50; n > 75; n > 100; n > 125; n > 150; n > 200; n > 225; n > 250; n > 350; n > 500; n > 600; n > 750; n > 1,000; n > 2,000; n > 4,000; n > 5,000; n > 7,500; n > 10,000; n > 20,000; n > 40,000; n > 75,000; n > 100,000; n > 200,000; n > 500,000, n > 1 x 106, n > 5 x 106, or n > 1 x 107. As such, in some embodiments, the algorithms, models, regressors, and / or classifiers of the present disclosure cannot be mentally performed. In some embodiments, n is between 20 and 100, 50 and 1000, and 100 and 5000, 10,000 and 1 x 107, between 100,000 and 5 x 106, or between 500,000 and 1 x 106. In some embodiments, the algorithms, models, regressors, and / or classifier of the present disclosure operate in a k-dimensional space, where k is a positive integer of 5 or greater (e.g., 5, 6, 7, 8, 9, 10, etc.). As such, the algorithms, models, regressors, and / or classifiers of the present disclosure cannot be mentally performed.

[0079] Neural networks. In some embodiments, a model used in the present disclosure is a neural network (e.g., a convolutional neural network and / or a residual neural network). Neural network models, also known as artificial neural networks (ANNs), include convolutional and / or residual neural network models (deep learning models). Neural networks can be machine learning models that may be trained to map an input data set to an output data set, where the neural network comprises an interconnected group of nodes organized into multiple layers of nodes. For example, the neural network architecture may comprise at least an input layer, one or more hidden layers, and an output layer. The neuralnetwork may comprise any total number of layers, and any number of hidden layers, where the hidden layers function as trainable feature extractors that allow mapping of a set of input data to an output value or set of output values. As used herein, a deep learning model (DNN) can be a neural network comprising a plurality of hidden layers, e.g., two or more hidden layers. Each layer of the neural network can comprise a number of nodes (or “neurons”). A node can receive input that comes either directly from the input data or the output of nodes in previous layers, and perform a specific operation, e.g., a summation operation. In some embodiments, a connection from an input to a node is associated with a parameter (e.g., a weight and / or weighting factor). In some embodiments, the node may sum up the products of all pairs of inputs, xi, and their associated parameters. In some embodiments, the weighted sum is offset with a bias, b. In some embodiments, the output of a node or neuron may be gated using a threshold or activation function, f, which may be a linear or non-linear function. The activation function may be, for example, a rectified linear unit (ReLU) activation function, a Leaky ReLU activation function, or other function such as a saturating hyperbolic tangent, identity, binary step, logistic, arcTan, softsign, parametric rectified linear unit, exponential linear unit, softPlus, bent identity, softExponential, Sinusoid, Sine, Gaussian, or sigmoid function, or any combination thereof.

[0080] The weighting factors, bias values, and threshold values, or other computational parameters of the neural network, may be “taught” or “learned” in a training phase using one or more sets of training data. For example, the parameters may be trained using the input data from a training data set and a gradient descent or backward propagation method so that the output value(s) that the ANN computes are consistent with the examples included in the training data set. The parameters may be obtained from a back propagation neural network training process.

[0081] Any of a variety of neural networks may be suitable for use in the present disclosure. Examples can include, but are not limited to, feedforward neural networks, radial basis function networks, recurrent neural networks, residual neural networks, convolutional neural networks, residual convolutional neural networks, and the like, or any combination thereof. In some embodiments, the machine learning makes use of a pre-trained and / or transfer-learned ANN or deep learning architecture. Convolutional and / or residual neural networks can be used as models in the present disclosure.

[0082] For instance, a deep neural network model comprises an input layer, a plurality of individually parameterized (e.g., weighted) convolutional layers, and an output scorer. The parameters (e.g., weights) of each of the convolutional layers as well as the input layercontribute to the plurality of parameters (e.g., weights) associated with the deep neural network model. In some embodiments, at least 100 parameters, at least 1000 parameters, at least 2000 parameters or at least 5000 parameters are associated with the deep neural network model. As such, deep neural network models require a computer to be used because they cannot be mentally solved. In other words, given an input to the model, the model output needs to be determined using a computer rather than mentally in such embodiments. See, for example, Krizhevsky et al., 2012, “Imagenet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Systems 2, Pereira, Burges, Bottou, Weinberger, eds., pp. 1097-1105, Curran Associates, Inc.; Zeiler, 2012 “ADADELTA: an adaptive learning rate method,”' CoRR, vol. abs / 1212.5701; and Rumelhart et al., 1988, “Neurocomputing: Foundations of research,” ch. Learning Representations by Back- propagating Errors, pp. 696-699, Cambridge, MA, USA: MIT Press, each of which is hereby incorporated by reference.

[0083] Neural network models, including convolutional neural network models, suitable for use as models are disclosed in, for example, Vincent et al., 2010, “Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion,” J Mach Learn Res 11, pp. 3371-3408; Larochelle et al., 2009, “Exploring strategies for training deep neural networks,” J Mach Learn Res 10, pp. 1-40; and Hassoun, 1995, Fundamentals of Artificial Neural Networks, Massachusetts Institute of Technology, each of which is hereby incorporated by reference. Additional example neural networks suitable for use as models are disclosed in Duda et al., 2001, Pattern Classification, Second Edition, John Wiley & Sons, Inc., New York; and Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York, each of which is hereby incorporated by reference in its entirety. Additional example neural networks suitable for use as models are also described in Draghici, 2003, Data Analysis Tools for DNA Microarrays, Chapman & Hall / CRC; and Mount, 2001, Bioinformatics: sequence and genome analysis, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, each of which is hereby incorporated by reference in its entirety.

[0084] In some embodiments, the neural network is a fully connected neural network with ReLU activation. For instance, in some embodiments, the model is a neural network comprising a corresponding one or more inputs, where each input in the corresponding one or more inputs is for a chemical structure for a test chemical compound, a corresponding first hidden layer comprising a corresponding plurality of hidden neurons, where each hidden neuron in the corresponding plurality of hidden neurons (i) is fully connected to each input inthe plurality of inputs, (ii) is associated with a first activation function type, and (iii) is associated with a corresponding parameter (e.g., weight) in a plurality of parameters for the neural network, and one or more corresponding neural network outputs, where each respective neural network output in the corresponding one or more neural network outputs (i) directly or indirectly receives, as input, an output of each hidden neuron in the corresponding plurality of hidden neurons, and (ii) is associated with a second activation function type. In some such embodiments, the neural network is a fully connected network.

[0085] In some embodiments, the neural network comprises a plurality of hidden layers. As described above, hidden layers are located between input and output layers (e.g., to capture additional complexity). In some embodiments, where there is a plurality of hidden layers, each hidden layer may have a same or a different respective number of neurons.

[0086] In some embodiments, each hidden neuron (e.g., in a respective hidden layer in a neural network) is associated with an activation function that performs a function on the input data (e.g., a linear or non-linear function). Generally, the purpose of the activation function is to introduce nonlinearity into the data such that the neural network is trained on representations of the original data and can subsequently “fit” or generate additional representations of new (e.g., previously unseen) data. Selection of activation functions (e.g., a first and / or a second activation function) is dependent on the use case of the neural network, as certain activation functions can lead to saturation at the extreme ends of a dataset (e.g., tanh and / or sigmoid functions). For instance, in some embodiments, an activation function (e.g., a first and / or a second activation function) is selected from any suitable activation functions known in the art, including but not limited to any activation function disclosed herein.

[0087] In some embodiments, each hidden neuron is further associated with a parameter (e.g., a weight and / or a bias value) that contributes to the output of the neural network, determined based on the activation function. In some embodiments, the hidden neuron is initialized with arbitrary parameters (e.g., randomized weights). In some alternative embodiments, the hidden neuron is initialized with a predetermined set of parameters.

[0088] In some embodiments, the plurality of hidden neurons in a neural network (e.g., across one or more hidden layers) is at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 16, at least 17, at least 18, at least 19, at least 20, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, or at least 500 neurons. In some embodiments, the plurality of hidden neurons is atleast 100, at least 500, at least 800, at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, at least 6000, at least 7000, at least 8000, at least 9000, at least 10,000, at least 15,000, at least 20,000, or at least 30,000 neurons. In some embodiments, the plurality of hidden neurons is no more than 30,000, no more than 20,000, no more than 15,000, no more than 10,000, no more than 9000, no more than 8000, no more than 7000, no more than 6000, no more than 5000, no more than 4000, no more than 3000, no more than 2000, no more than 1000, no more than 900, no more than 800, no more than 700, no more than 600, no more than 500, no more than 400, no more than 300, no more than 200, no more than 100, or no more than 50 neurons. In some embodiments, the plurality of hidden neurons is from 2 to 20, from 2 to 200, from 2 to 1000, from 10 to 50, from 10 to 200, from 20 to 500, from 100 to 800, from 50 to 1000, from 500 to 2000, from 1000 to 5000, from 5000 to 10,000, from 10,000 to 15,000, from 15,000 to 20,000, or from 20,000 to 30,000 neurons. In some embodiments, the plurality of hidden neurons falls within another range starting no lower than 2 neurons and ending no higher than 30,000 neurons.

[0089] In some embodiments, the neural network comprises from 1 to 50 hidden layers. In some embodiments, the neural network comprises from 1 to 20 hidden layers. In some embodiments, the neural network comprises at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 13, at least 14, at least 15, at least 16, at least 17, at least 18, at least 19, at least 20, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, or at least 100 hidden layers. In some embodiments, the neural network comprises no more than 100, no more than 90, no more than 80, no more than 70, no more than 60, no more than 50, no more than 40, no more than 30, no more than 20, no more than 10, no more than 9, no more than 8, no more than 7, no more than 6, or no more than 5 hidden layers. In some embodiments, the neural network comprises from 1 to 5, from 1 to 10, from 1 to 20, from 10 to 50, from 2 to 80, from 5 to 100, from 10 to 100, from 50 to 100, or from 3 to 30 hidden layers. In some embodiments, the neural network comprises a plurality of hidden layers that falls within another range starting no lower than 1 layer and ending no higher than 100 layers.

[0090] In some embodiments, the neural network comprises a shallow neural network. A shallow neural network refers to a neural network with a small number of hidden layers. In some embodiments, such neural network architectures improve the efficiency of neural network training and conserve computational power due to the reduced number of layers involved in the training. In some embodiments, the neural network comprises one hiddenlayer. In some embodiments, the neural network comprises two, three, four, or five hidden layers.

[0091] In some embodiments, the neural network is a message passing neural network. A message passing neural network refers to a framework for supervised learning on graphs (e.g., graph-based representations of chemical structures), where nodes represent atoms and edges represent bonds between atoms. Generally, a message passing neural network comprises two phases in the forward pass, a message passing phase and a readout phase. The message passing phase runs for a period of / "intervals and comprises updating hidden states at each node in the graph in accordance with message functions Mt and vertex update functions Ut. The readout phase computes a feature vector for the graph using a readout function R. In some embodiments, the message passing neural network comprises a convolutional network (e.g., a spatial graph convolutional network and / or a spectral graph convolutional network), a gated graph neural network (GG-NN), an interaction network, a molecular graph convolution, a deep tensor neural network, and / or a Laplacian-based method. See, for example, Gilmer et al., 2017, “Neural Message Passing for Quantum Chemistry,” arXiv: 1704.01212v2, which is hereby incorporated herein by reference in its entirety.

[0092] Support vector machines. In some embodiments, a model used in the present disclosure is a support vector machine (SVM). Suitable SVM models are described in, for example, Cristianini and Shawe-Taylor, 2000, “An Introduction to Support Vector Machines,” Cambridge University Press, Cambridge; Boser et al., 1992, “A training algorithm for optimal margin models,” in Proceedings of the 5th Annual ACM Workshop on Computational Learning Theory, ACM Press, Pittsburgh, Pa., pp. 142-152; Vapnik, 1998, Statistical Learning Theory, Wiley, New York; Mount, 2001, Bioinformatics: sequence and genome analysis, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.; Duda, Pattern Classification, Second Edition, 2001, John Wiley & Sons, Inc., pp. 259, 262-265; and Hastie, 2001, The Elements of Statistical Learning, Springer, New York; and Furey et al., 2000, Bioinformatics 16, 906-914, each of which is hereby incorporated by reference in its entirety. When used for classification, SVMs separate a given set of binary labeled data with a hyper-plane that is maximally distant from the labeled data. For cases in which no linear separation is possible, SVMs can work in combination with the technique of 'kernels', which automatically realizes a non-linear mapping to a feature space. The hyper-plane found by the SVM in feature space can correspond to a non-linear decision boundary in the input space. In some embodiments, the plurality of parameters (e.g., weights) associated with the SVMdefine the hyper-plane. In some embodiments, the hyper-plane is defined by at least 10, at least 20, at least 50, or at least 100 parameters and the SVM model requires a computer to calculate because it cannot be mentally solved.

[0093] Naive Bayes models. In some embodiments, a model used in the present disclosure is a Naive Bayes model. Naive Bayes models suitable for use as models are disclosed, for example, in Ng etal., 2002, “On discriminative vs. generative classifiers: A comparison of logistic regression and naive Bayes,” Advances in Neural Information Processing Systems, 14, which is hereby incorporated by reference. A Naive Bayes classifier is any classifier in a family of “probabilistic classifiers” based on applying Bayes’ theorem with strong (naive) independence assumptions between the features. In some embodiments, they are coupled with Kernel density estimation. See, for example, Hastie et al., 2001, The elements of statistical learning : data mining, inference, and prediction, eds. Tibshirani and Friedman, Springer, New York, which is hereby incorporated by reference.

[0094] Nearest neighbor models. In some embodiments, a model used in the present disclosure is a nearest neighbor model. Nearest neighbor models can be memory-based and include no model to be fit. For nearest neighbors, given a query point xo (a test subject), the k training points X(r), r, ... , k (here the training subjects) closest in distance to xo are identified and then the point xo is classified using the k nearest neighbors. Here, the distance to these neighbors is a function of the abundance values of the discriminating gene set. In some embodiments, Euclidean distance in feature space is used to determine distance as=■ Typically, when the nearest neighbor model is used, the abundance data used to compute the linear discriminant is standardized to have mean zero and variance 1. The nearest neighbor rule can be refined to address issues of unequal class priors, differential misclassification costs, and feature selection. Many of these refinements involve some form of weighted voting for the neighbors. For more information on nearest neighbor analysis, see Duda, Pattern Classification, Second Edition, 2001, John Wiley & Sons, Inc; and Hastie, 2001, The Elements of Statistical Learning, Springer, New York, each of which is hereby incorporated by reference.

[0095] A k-nearest neighbor model is a non-parametric machine learning method in which the input consists of the k closest training examples in feature space. The output is a class membership. An object is classified by a plurality vote of its neighbors, with the object being assigned to the class most common among its k nearest neighbors (k is a positive integer, typically small). If k = 1, then the object is simply assigned to the class of that single nearest neighbor. See, Duda et al., 2001, Pattern Classification, Second Edition, John Wiley& Sons, which is hereby incorporated by reference. In some embodiments, the number of distance calculations needed to solve the k-nearest neighbor model is such that a computer is used to solve the model for a given input because it cannot be mentally performed.

[0096] Random forest, decision tree, and boosted tree models. In some embodiments, a model used in the present disclosure is a decision tree. Decision trees suitable for use as models are described generally by Duda, 2001, Pattern Classification, John Wiley & Sons, Inc., New York, pp. 395-396, which is hereby incorporated by reference. Tree-based methods partition the feature space into a set of rectangles, and then fit a model (like a constant) in each one. In some embodiments, the decision tree is random forest regression. One specific model that can be used is a classification and regression tree (CART). Other specific decision tree models include, but are not limited to, ID3, C4.5, MART, and Random Forests. CART, ID3, and C4.5 are described in Duda, 2001, Pattern Classification, John Wiley & Sons, Inc., New York, pp. 396-408 and pp. 411-412, which is hereby incorporated by reference. CART, MART, and C4.5 are described in Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York, Chapter 9, which is hereby incorporated by reference in its entirety. Random Forests are described in Breiman, 1999, “Random Forests— Random Features,” Technical Report 567, Statistics Department, U.C. Berkeley, September 1999, which is hereby incorporated by reference in its entirety. In some embodiments, the decision tree model includes at least 10, at least 20, at least 50, or at least 100 parameters (e.g., weights and / or decisions) and requires a computer to calculate because it cannot be mentally solved.

[0097] Regression. In some embodiments, a model used in the present disclosure is a regression algorithm. The regression algorithm can be any type of regression. For example, in some embodiments, the regression is logistic regression. In some embodiments, the regression is logistic regression with lasso, L2 or elastic net regularization. In some embodiments, the regression is logistic regression with LI elastic net regularization. In some embodiments, those extracted features that have a corresponding regression coefficient that fails to satisfy a threshold value are pruned (removed from) consideration. In some embodiments, a generalization of the logistic regression model that handles multicategory responses is used as the model. Logistic regression is disclosed in Agresti, An Introduction to Categorical Data Analysis, 1996, Chapter 5, pp. 103-144, John Wiley & Son, New York, which is hereby incorporated by reference. In some embodiments, the model makes use of a regression model disclosed in Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York. In some embodiments, the logistic regression model includes atleast 10, at least 20, at least 50, at least 100, or at least 1000 parameters (e.g., weights). In some embodiments the logistic regression model requires a computer to calculate because it cannot be mentally performed.

[0098] Linear discriminant analysis. Linear discriminant analysis (LDA), normal discriminant analysis (ND A), or discriminant function analysis can be a generalization of Fisher’s linear discriminant, a method used in statistics, pattern recognition, and machine learning to find a linear combination of features that characterizes or separates two or more classes of objects or events. The resulting combination can be used as a model (linear model) in some embodiments of the present disclosure.

[0099] Mixture model and Hidden Markov model. In some embodiments, a model used in the present disclosure is a mixture model, such as that described in McLachlan et al., Bioinformatics 18(3):413-422, 2002. In some embodiments, in particular, those embodiments including a temporal component, the model is a hidden Markov model such as described by Schliep etal., 2003, Bioinformatics 19(l):i255-i263.

[0100] Ensembles of models and boosting. In some embodiments, an ensemble (two or more) of models is used. In some embodiments, a boosting technique such as AdaBoost is used in conjunction with many other types of learning algorithms to improve the performance of the model. In this approach, the output of any of the models disclosed herein, or their equivalents, is combined into a weighted sum that represents the final output of the boosted model. In some embodiments, the plurality of outputs from the models is combined using any measure of central tendency known in the art, including but not limited to a mean, median, mode, a weighted mean, weighted median, weighted mode, etc. In some embodiments, the plurality of outputs is combined using a voting method. In some embodiments, a respective model in the ensemble of models is weighted or unweighted.

[0101] I. Exemplary System Embodiments

[0102] Now that an overview of some aspects of the present disclosure and some definitions used in the present disclosure have been provided, details of an exemplary system are described in conjunction with Figure 1.

[0103] Figure 1 provides a block diagram illustrating a system 100 in accordance with some embodiments of the present disclosure. The system 100 assesses a risk that a test chemical compound will induce a liver injury in a human subject. In Figure 1, the system 100 is illustrated as a computing device. Other topologies of the computer system 100 are possible. For instance, in some embodiments, the system 100 can in fact constitute several computer systems that are linked together in a network, or be a virtual machine or a containerin a cloud computing environment. As such, the exemplary topology shown in Figure 1 merely serves to describe the features of an embodiment of the present disclosure in a manner that will be readily understood to one of skill in the art.

[0104] Referring to Figure 1, in some embodiments a computer system 100 (e.g., a computing device) includes a network interface 104. In some embodiments, the network interface 104 interconnects the system 100 computing devices within the system with each other, as well as optional external systems and devices, through one or more communication networks. In some embodiments, the network interface 104 optionally provides communication via the Internet, one or more local area networks (LANs), one or more wide area networks (WANs), other types of networks, or a combination of such networks.

[0105] Examples of networks include the World Wide Web (WWW), an intranet and / or a wireless network, such as a cellular telephone network, a wireless local area network (LAN) and / or a metropolitan area network (MAN), and other devices by wireless communication. The wireless communication optionally uses any of a plurality of communications standards, protocols and technologies, including Global System for Mobile Communications (GSM), Enhanced Data GSM Environment (EDGE), high-speed downlink packet access (HSDPA), high-speed uplink packet access (HSUPA), Evolution, Data-Only (EV-DO), HSPA, HSPA+, Dual-Cell HSPA (DC-HSPDA), long term evolution (LTE), near field communication (NFC), wideband code division multiple access (W-CDMA), code division multiple access (CDMA), time division multiple access (TDMA), Bluetooth, Wireless Fidelity (Wi-Fi) (e.g., IEEE 802. I la, IEEE 802.1 lac, IEEE 802.1 lax, IEEE 802.1 lb, IEEE 802.11g and / or IEEE 802.1 In), voice over Internet Protocol (VoIP), Wi-MAX, a protocol for e-mail (e.g., Internet message access protocol (IMAP) and / or post office protocol (POP)), instant messaging (e.g., extensible messaging and presence protocol (XMPP), Session Initiation Protocol for Instant Messaging and Presence Leveraging Extensions (SIMPLE), Instant Messaging and Presence Service (IMPS)), and / or Short Message Service (SMS), or any other suitable communication protocol, including communication protocols not yet developed as of the filing date of this document.

[0106] The system 100 in some embodiments includes one or more processing units (CPU(s)) 102 (e.g., a processor, a processing core, etc. , one or more network interfaces 104, a user interface 106 including (optionally) a display 108 and an input system 105 (e.g., an input / output interface, a keyboard, a mouse, etc.) for use by the user, memory (e.g., non- persistent memory 107, persistent memory), and one or more communication buses 103 for interconnecting the aforementioned components. The one or more communication buses 103optionally include circuitry (sometimes called a chipset) that interconnects and controls communications between system components. The non-persistent memory 107 typically includes high-speed random access memory, such as DRAM, SRAM, DDR RAM, ROM, EEPROM, flash memory, whereas persistent memory (not shown) typically includes CD- ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, magnetic disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid state storage devices. The persistent memory 109 optionally includes one or more storage devices remotely located from the CPU(s) 102. The persistent memory, and the non-volatile memory device(s) within the non-persistent memory, include non-transitory computer readable storage medium. In some embodiments, the non-persistent memory 107 or alternatively the non- transitory computer readable storage medium stores the following programs, modules and data structures, or a subset thereof, sometimes in conjunction with the persistent memory:• an optional operating system (not shown) (e.g., ANDROID, iOS, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks), which includes procedures for handling various basic system services and for performing hardware dependent tasks;• an optional network communication module (not shown) (or instructions) for connecting the system 100 with other devices and / or a communication network 104;• a cellular constituent dataset 114 comprising, for a test chemical compound 114, for each concentration 116 / exposure time 118 pair (e.g., 116-1 / 118-1 to 116-M / l 18-M, where M is a positive integer of two or greater), a measured cellular constituent abundance value 122 (e.g., 122-1-1, 122-1-2, ..., 122-1-N, where N is a positive integer of 3 or greater) for each cellular constituent 120 (e.g., 120-1-1, 120-1-2, . . ., 120-1-N);• one or more first models 130 for assessing DILI risk comprising a plurality of model parameters 132 (e.g., 132-1, . . ., 132-K, where K is a positive integer) and configured to output one or more first model outputs 134;• one or more second models 136 (e.g, second models 136-1, 136-2, ..., 136-T) for assessing DILI risk, each model 136 comprising a corresponding plurality of model parameters 138 (e.g., 138-1-1, ..., 138-1-Q, where Q is a positive integer) and configured to output a second model 2 output 140 (e.g., 140-1); and• a plurality of pathway modules 142 (e.g, 142-1, 142-2, ..., 142-X), each pathway module associated with a subset of cellular constituents 144 (e.g., 144-1-1, 144-1-2,144-1-P, where P is a positive integer; 144-2-1, 144-2-2, 144-2-W, where W is a positive integer, 144-X-l, 144-X-2, 144-X-R, where R is a positive integer).

[0107] In various embodiments, one or more of the above identified elements are stored in one or more of the previously mentioned memory devices and correspond to a set of instructions for performing a function described above. The above identified modules, data, or programs (e.g., sets of instructions) need not be implemented as separate software programs, procedures, datasets, or modules, and thus various subsets of these modules and data may be combined or otherwise re-arranged in various implementations. In some implementations, the non-persistent memory 107 optionally stores a subset of the modules and data structures identified above. Furthermore, in some embodiments, the memory stores additional modules and data structures not described above. In some embodiments, one or more of the above identified elements is stored in a computer system, other than that of the system 100, that is addressable by the system 100 so that the system 100 may retrieve all or a portion of such data when needed.

[0108] Although Figure 1 depicts a “system 100,” the figure is intended more as a functional description of the various features that may be present in computer systems than as a structural schematic of the implementations described herein. In practice, and as recognized by those of ordinary skill in the art, items shown separately could be combined and some items could be separated. Moreover, although Figure 1 depicts certain data and modules in non-persistent memory 107, some or all of these data and modules instead may be stored in persistent memory or in more than one memory. For example, in some embodiments, at least the cellular constituent dataset 112 can be stored in a remote storage device that can be a part of a cloud-based infrastructure.

[0109] IL Methods of Assessing a Risk that a test Chemical Compound Will Induce a Liver Injury in a Subject

[0110] Referring to block 200, in some embodiments, a method of assessing a risk that a test chemical compound will induce a liver injury in a human subject is provided.

[0111] Block 202. Referring to block 202, in some embodiments, the test chemical compound is an organic compound having a molecular weight of less than 2000 Daltons. In some embodiments, the test chemical compound is inorganic or organic. In some embodiments, the test chemical compound is an organic compound having a molecular weight of less than 2000 Daltons (Da). In some embodiments, the test chemical compound has a molecular weight of at least 10 Da, at least 20 Da, at least 50 Da, at least 100 Da, at least 200 Da, at least 500 Da, at least 1 kDa, at least 2 kDa, at least 3 kDa, at least 5 kDa, atleast 10 kDa, at least 20 kDa, at least 30 kDa, at least 50 kDa, at least 100 kDa, or at least 500 kDa. In some embodiments, the test chemical compound has a molecular weight of no more than 1000 kDa, no more than 500 kDa, no more than 100 kDa, no more than 50 kDa, no more than 10 kDa, no more than 5 kDa, no more than 2 kDa, no more than 1 kDa, no more than 500 Da, no more than 300 Da, no more than 100 Da, or no more than 50 Da. In some embodiments, the test chemical compound has a molecular weight of from 10 Da to 900 Da, from 50 Da to 1000 Da, from 100 Da to 2000 Da, from 1 kDa to 10 kDa, from 5 kDa to 500 kDa, or from 100 kDa to 1000 kDa. In some embodiments, the test chemical compound has a molecular weight that falls within another range starting no lower than 10 Daltons and ending no higher than 1000 kDa.

[0112] Blocks 204 and 206. Referring to block 204, in some embodiments, the test chemical compound is an organic compound that satisfies each of the Lipinski rule of five criteria. Referring to block 206, in some embodiments, the test chemical compound is an organic compound that satisfies at least three criteria of the Lipinski rule of five criteria. In some embodiments, the test chemical compound is an organic compound that satisfies two or more rules, three or more rules, or all four rules of the Lipinski's Rule of Five: (i) not more than five hydrogen bond donors (e.g., OH and NH groups), (ii) not more than ten hydrogen bond acceptors (e.g. N and O), (iii) a molecular weight under 500 Daltons, and (iv) a LogP under 5. The “Rule of Five” is so called because three of the four criteria involve the number five. See, Lipinski, 1997, Adv. Drug Del. Rev. 23, 3, which is hereby incorporated herein by reference in its entirety. In some embodiments, a compound of the present disclosure satisfies one or more criteria in addition to Lipinski's Rule of Five. For example, in some embodiments, a compound of the present disclosure has five or fewer aromatic rings, four or fewer aromatic rings, three or fewer aromatic rings, or two or fewer aromatic rings.

[0113] Referring to block 208, in some embodiments, the liver injury is hepatitis, hepatotoxicityjaundice, hepatic fibrosis, cirrhosis, hepatomegaly, alcoholic and nonalcoholic Liver fatty Liver disease and / or choluria.

[0114] Drug induced liver disease can be classified into three injury patterns: hepatocellular, cholestatic, and mixed hepatocellular-cholestatic. These designations refer to histologic features of injury, but are usually defined based upon the pattern of serum enzyme elevations. See, LiverTox: Clinical and Research Information on Drug-Induced Liver Injury [Internet], Bethesda (MD): National Institute of Diabetes and Digestive and Kidney Diseases; 2012-. Clinical Course and Diagnosis of Drug Induced Liver Disease. [Updated 2019 May 4],Available from: h p / / wwwngbimlm. ^ which is hereby incorporated by reference.

[0115] Hepatitis, hepatotoxicity. Drug induced liver disease that resembles acute viral hepatitis is typified by a prominent hepatocellular pattern of injury. Liver biopsy, if available, usually shows marked liver cell necrosis and inflammation with only mild bile stasis, at least in the early stages. If present, symptoms of fatigue and weakness predominate. Serum alanine and aspartate aminotransferase (ALT and AST) levels typically are markedly elevated (usually > 10-fold), while the alkaline phosphatase or gamma glutamyl transpeptidase (GGT) are only modestly increased. An “R” ratio of ALT to alkaline phosphatase (both expressed as multiples of the upper limit of the normal range) of 5 or more is often used to define a hepatocellular pattern of injury. Id.

[0116] Cholestatic injury. A cholestatic picture of drug induced liver injury resembles bile duct obstruction or choledocholithiasis. The liver biopsy findings are generally of bile stasis, portal inflammation and proliferation or injury of bile ducts and ductules. Clinically, symptoms of jaundice and itching predominate. Id.

[0117] Mixed hepatocellular-cholestatic injury. A mixture of hepatocellular and cholestatic injury is typical of many drugs and, indeed, is the pattern that is most characteristic of drug induced liver injury, occurring rarely in other forms of acute liver disease. In cases of mixed injury, liver biopsy shows prominent hepatocyte necrosis and inflammation accompanied by marked bile stasis. Symptoms may include both fatigue and itching, and laboratory tests show similar elevations in serum ALT and alkaline phosphatase. An R ratio of ALT to alkaline phosphatase (both expressed as multiples of the upper limit of the normal range) between 2 and 5 is used to define a mixed pattern of injury. Drugs that cause a mixed hepatocellular-cholestatic pattern of injury include the sulfonamides, phenytoin and enalopril. Id.

[0118] In addition, to the injury patterns, drug induced liver injury can be categorized by an overall clinical pattern and course of injury into at least twelve “phenotypes”. These phenotypes overlap to some degree with the pattern of injury (hepatocellular, cholestatic, mixed) and include: acute hepatic necrosis, acute (hepatocellular) hepatitis, cholestatic hepatitis, mixed hepatitis, serum enzyme elevations without jaundice, bland cholestasis, acute fatty liver with lactic acidosis, nonalcoholic fatty liver, chronic hepatitis, sinusoidal obstruction syndrome, nodular regenerative hyperplasia, and liver tumors, such as hepatic adenoma and hepatocellular carcinoma. In addition, characteristic features or outcomes areoften assigned to each phenotype such as immunoallergic features, autoimmune features, acute liver failure, vanishing bile duct syndrome and cirrhosis. Id.

[0119] Referring to block 210, in some embodiments, the liver injury is characterized by an elevation of one or more liver enzymes, cholestasis, and / or alcoholic stools. For example, in some embodiments, serum alanine and aspartate aminotransferase (ALT and AST) levels are elevated more than 10-fold, while the alkaline phosphatase or gamma glutamyl transpeptidase (GGT) are modestly increased. In some embodiments liver injury is characterized by a liver function tests, such as ALP (alkaline phosphatase), ALT (alanine transaminase), AST (aspartate aminotransferase), or gamma-glutamyl tansferase (GGT). In some embodiments liver injury is characterized by a Histology (e.g., a liver biopsy). In some embodiments liver injury is characterized by transient elastrography (Fibroscan). See for example Foucher etal., 2006, “Diagnosis of cirrhosis by transient elastography (FibroScan): a prospective study,” Gut 55(3), pp. 403-408, which is hereby incorporated by reference. In some embodiments liver injury is characterized by imaging such as ultrasound, computed tomography (CT) scan, or magnetic resonance imaging (MRI). In some embodiments, liver injury is characterized by a Child-Plugh score to determine Cirrhosis (e.g., using five clinical parameters: bilirubin, albumin, INR, ascites, hepatic, encephalopathy). See for example, Tsoris A, Marlar CA., “Use Of The Child Pugh Score In Liver Disease,” [Updated 2023 Mar 13], In: StatPearls [Internet], Treasure Island (FL): StatPearls Publishing; 2023 Jan-.Available on the Internet at ncbi.nlm.nih.gov / books / NBK542308. In some embodiments, liver injury is characterized by viral load (e.g., viral hepatitis). See for example, Oliveira et al., “Advanced Liver Injury in Patients with Chronic Hepatitis B And Viral Load Below 2,000 lu / Ml,” Rev Inst Med Trop Sao Paulo. 2016 Sep 22;58:65. doi: 10.1590 / S 1678- 9946201658065. PMID: 27680170; PMCID: PMC5048636, which is hereby incorporated by reference.

[0120] In cholestasis without or with only modest liver cell injury itching and jaundice are prominent, but, otherwise, patients typically feel well. Serum alkaline phosphatase and ALT levels may be normal or minimally (less that two-fold) elevated, particularly at the peak of symptoms and jaundice. At the onset, ALT levels may be more prominently elevated (3 to 10 fold) and in more severe instances, serum alkaline phosphatase may rise above 2 fold. Pure or bland cholestasis is the characteristic form of liver injury caused by estrogenic and anabolic steroids and can occur with illicit use of muscle building regimens and more rarely with the thioguanines such as azathioprine and mercaptopurine. Id.

[0121] Blocks 212-214. Referring to block 212, in some embodiments, a cellular constituent dataset 112 is obtained. In some embodiments the cellular constituent dataset 112 is in electronic form. The dataset comprises a plurality of cellular constituent abundance values for a plurality of cellular constituents in a first plurality of cells that have been exposed to the test chemical compound for a first period of time. In some embodiments, the plurality of cellular constituents is measured at a single time point. In some embodiments, the plurality of cellular constituents is measured at multiple time points. In embodiments where the plurality of cellular constituents is measured at multiple time points, a first subset of the plurality of cells is exposed to the test chemical compound for a first period of time (118-1), a second subset of the plurality of cells is exposed to the test chemical compound for a second period of time (118-2), . . ., and an M* subset of cells are exposed to the test chemical compound for an M* period of time, where M is a positive integer. In some embodiments, M is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 or greater, 20 or greater or 30 or greater. For each of these M subsets, the exposure time may be the same or different. For instance, in some embodiments, the exposure time is different, for instance ranging from 4 hours for one of the time points, to 24 hours for another of the time points.

[0122] In some embodiments each cellular constituent value in the plurality of cellular constituent abundance values for the plurality of cellular constituents in the first plurality of cells that have been exposed to the test chemical compound for a first period of time is in the form of cellular constituent expression log-fold change (DMSO vs. compound perturbation). For example in some embodiments the cellular constituent are genes and the gene expression readout measured is measured by RNA-seq, in order to get transcriptome abundance (e.g., by SMART-Seq DE) from perturbations (cells exposed to the test composition at a particular concentration) and DMSO (cells exposed to DMSO only). In some embodiments the log-fold changes are determined using a program such as DESeq2, (htps: / / gen mebiology.bioiuedcentr ..com / an.ides / I0 1J.86 / s 1.3059^)1 -0550-8), Love et al., 2014, “Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2,” Genome Biol 15, 550 on the Internet at doi.org / 10.1186 / sl3059-014-0550-8 which normalizes and provides log-fold change estimation of perturbation profiles (cells exposed to test compound) against DMSO (cells exposed to DMSO only). DESeq2 first estimates size factors for each sample. The size factor represents the scaling factor that should be applied to each sample to account for differences in sequencing depth. DESeq2 computes this as the median of the ratios of gene counts to their geometric means across all samples (e.g., three replicates, etc.). DESeq2 then uses these size factors to normalize the raw counts by dividingeach gene's count in a sample by the corresponding size factor for that sample. This normalization step adjusts the counts for sequencing depth, making the samples comparable, log-fold changes were then determined by taking the difference in the log2-transformed normalized counts (log2-counts) between perturbation (cells exposed to test compound at a given concentration for a given time) versus DMSO (cells exposed to DMSO for a given time) for each gene. DESeq2 employs a shrinkage procedure, called “regularized log transformation,” to moderate log-fold changes. This helps improve the accuracy of log-fold changes for genes with low counts and high dispersion. Finally, DESeq2 performed statistical tests (Wald test) to determine if the observed log-fold changes were significantly different from zero. If so, they were used for the log-fold changes that serve as the cellular constituent abundance values for the plurality of cellular constituents in the first plurality of cells that have been exposed to the test chemical compound for a first period of time.

[0123] In some embodiments, the exposure time for each of the M subsets is the same and what is varied is the concentration 116 at each of the M subsets. In some embodiments, each subset of the plurality of cells is exposed to a different concentration of the test chemical compound. For instance, in some embodiments a separate PK study is optionally used to estimate the Cmax of the test chemical compound. Then a range of exposure concentrations 116 is selected that at least encompasses the Cmax of the test chemical compound so that a predicted safety margin can be computed using the cellular constituent dataset. In some embodiments, the range of exposure concentrations is from 0.1 pM to 1 pM. In some embodiments, the range of exposure concentrations is from 0.01 pM to 1 pM. In some embodiments, the range of exposure concentrations is from 0.1 pM to 10 pM. In some embodiments, the lowest concentration used in the range of exposure concentrations is 0.001 pM, 0.01 pM, 0.01 pM, or 1 pM. In some embodiments, the highest concentration used in the range of exposure concentrations is 100 pM, 50 pM, 10 pM, 1 pM, 0.5 pM, or 0.1 pM. In some embodiments, the cellular constituent data provide a separate set of cellular constituent abundance values for the plurality of cellular constituents from 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 different subsets of the plurality of cells where each different subset of the plurality of cells was exposed to the test chemical compound at a different exposure concentration. In some embodiments, such measurements are performed in duplicate, triplicate or higher order redundancy. For instance, in some embodiments such measurements are performed in triplicate. As an example, for a given exposure concentration 116 / exposure time 118 pair, three different subsets of the plurality of cells are independently (e.g., in separate wells, separate aliquots of cells), exposed to the test chemical compositionfor the exposure time 118 at the given exposure concentration 116 and the cellular constituent abundance values of the plurality of cellular constituents is separately from each of the separate wells / separate aliquots of cells. For instance, the cellular constituent abundance values of each of the cellular constituents in the plurality of cellular constituents is measured in the cells in the first well / aliquot of cells, the cellular constituent abundance values of each of the cellular constituents in the plurality of cellular constituents is measured in the cells in the second well / aliquot of cells, and the cellular constituent abundance values of each of the cellular constituents in the plurality of cellular constituents is measured in the cells in the third well / aliquot of cells. These measurements values are then combined to arrive at the cellular constituent abundance values for the plurality of cellular constituents of the given exposure concentration 116 / exposure time 118 pair.

[0124] Referring to block 214, in some embodiments, each cellular constituent in the plurality of cellular constituents is a particular gene, a particular mRNA associated with a gene, a carbohydrate, a lipid, an epigenetic feature, a metabolite, a protein, or a combination thereof.

[0125] In some embodiments, a cellular constituent is a gene, a gene product (e.g., an mRNA and / or a protein), a carbohydrate, a lipid, an epigenetic feature, a metabolite, and / or a combination thereof. In some embodiments, each cellular constituent in the plurality of cellular constituents is a particular gene, a particular mRNA associated with a gene, a carbohydrate, a lipid, an epigenetic feature, a metabolite, a protein, or a combination thereof. In some embodiments, the plurality of cellular constituents includes nucleic acids, including DNA, modified (e.g., methylated) DNA, RNA, including coding (e.g., mRNAs) or noncoding RNA (e.g., sncRNAs), proteins, including post-transcriptionally modified protein (e.g., phosphorylated, glycosylated, myristilated, etc. proteins), lipids, carbohydrates, nucleotides (e.g., adenosine triphosphate (ATP), adenosine diphosphate (ADP) and adenosine monophosphate (AMP)) including cyclic nucleotides such as cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP), other small molecule cellular constituents such as oxidized and reduced forms of nicotinamide adenine dinucleotide (NADP / NADPH), and any combinations thereof.

[0126] In some embodiments, the first plurality of cells comprises at least 5, at least 10, at least 15, at least 20, at least 30, at least 40, at least 50, at least 100, at least 200, at least 300, at least 400, at least 500, at least 1000, at least at least 2000, at least 3000, at least 4000, at least 5000, at least 10,000, at least 20,000, at least 30,000, at least 50,000, at least 80,000, at least 100,000, at least 500,000, or at least 1 million cells. In some embodiments, the firstplurality of cells comprises no more than 5 million, no more than 1 million, no more than 500,000, no more than 100,000, no more than 50,000, no more than 10,000, no more than 5000, no more than 1000, no more than 500, no more than 200, no more than 100, or no more than 50 cells. In some embodiments, the first plurality of cells comprises from 5 to 100, from 10 to 50, from 20 to 500, from 200 to 10,000, from 1000 to 100,000, from 50,000 to 500,000, or from 10,000 to 1 million cells. In some embodiments, the first plurality of cells falls within another range starting no lower than 5 cells and ending no higher than 5 million cells.

[0127] Referring to block 216, in some embodiments, the first plurality of cells is primary human hepatocytes. In some embodiments the primary human hepatocytes are solubilized in a certain percentage of DMSO.

[0128] Blocks 218-224. Referring to block 218, in some embodiments, the plurality of cellular constituent abundance values is determined by single-cell ribonucleic acid (RNA) sequencing (scRNA-seq) data of the first plurality of cells.

[0129] Referring to block 220, in some embodiments, the plurality of cellular constituent abundance values is determined by bulk RNA sequencing. For example, in some embodiments, for a given concentration 116 / exposure time pair 118, the respective cellular constituent abundance value measured for a respective cellular constituent for each of the cells exposed to the given concentration 116 for the given exposure time is averaged together.

[0130] Referring to block 222, in some embodiments, the plurality of cellular constituent abundance values is determined by scTag-seq. Referring to block 224, in some embodiments, the plurality of cellular constituent abundance values is determined by singlecell assay for transposase-accessible chromatin using sequencing (scATAC-seq), CyTOF / SCoP, E-MS / Abseq, miRNA-seq, CITE-seq, or any combination thereof.

[0131] In some embodiments, the cellular constituent abundances for the plurality of cellular constituents in the first plurality of cells are preprocessed. In some embodiments, the preprocessing includes one or more of filtering, normalization, mapping (e.g. , to a reference sequence), quantification, scaling, deconvolution, cleaning, dimension reduction, transformation, statistical analysis, and / or aggregation.

[0132] For example, in some embodiments, the plurality of cellular constituents is filtered based on a desired quality, e.g., size and / or quality of a nucleic acid sequence, or a minimum and / or maximum abundance value for a respective cellular constituent. In some embodiments, filtering is performed in part or in its entirety by various software tools, such as Skewer. See, Jiang, H. et al., BMC Bioinformatics 15(182): 1-12 (2014). In some embodiments, the plurality of cellular constituents is filtered for quality control, for example,using a sequencing data QC software such as AfterQC, Kraken, RNA-SeQC, FastQC, or another similar software program. In some embodiments, the plurality of cellular constituents is normalized, e.g., to account for pull-down, amplification, and / or sequencing bias (e.g., mappability, GC bias etc.). See, for example, Schwartz et al., PLoS ONE 6(l):el6685 (2011) and Benjamini and Speed, Nucleic Acids Research 40(10):e72 (2012), the contents of which are hereby incorporated by reference, in their entireties, for all purposes. In some embodiments, the preprocessing removes a subset of cellular constituents from the first plurality of cellular constituents. In some embodiments, the preprocessing the corresponding abundances for the first plurality of cellular constituents improves (e.g., lowers) a high signal-to-noise ratio.

[0133] Thus, in some embodiments, the corresponding abundance of a respective cellular constituent in the cellular constituent dataset comprises any one of a variety of forms, including, without limitation, a raw abundance value, an absolute abundance value (e.g., transcript number), a relative abundance value (e.g., relative fluorescent units, transcriptome analysis, and / or gene set expression analysis (GSEA)), a compound or aggregated abundance value, a transformed abundance value (e.g., Iog2 and / or log transformed), a change (e.g., fold- or log-change) relative to a reference (e.g., a normal sample, matched sample, reference dataset, housekeeping gene, and / or reference standard), a standardized abundance value, a measure of central tendency (e.g., mean, median, mode, weighted mean, weighted median, and / or weighted mode), a measure of dispersion (c.g, variance, standard deviation, and / or standard error), or an adjusted abundance value (e.g., normalized, scaled, and / or error- corrected).

[0134] Any one of a number of abundance counting techniques (e.g., cellular constituent measurement techniques) may be used to obtain the abundance value of each of the cellular constituents in the cellular constituent dataset.

[0135] In some embodiments, the abundance of each cellular constituent in the plurality of cellular constituents of cellular constituent dataset is determined using one or more methods including microarray analysis via fluorescence, chemiluminescence, electric signal detection, polymerase chain reaction (PCR), reverse transcriptase polymerase chain reaction (RT-PCR), digital droplet PCR (ddPCR), solid-state nanopore detection, RNA switch activation, a Northern blot, and / or a serial analysis of gene expression (SAGE).

[0136] In some embodiments, the abundance of each cellular constituent in the plurality of cellular constituents of cellular constituent dataset is determined by a colorimetricmeasurement, a fluorescence measurement, a luminescence measurement, or a resonance energy transfer (FRET) measurement.

[0137] In some embodiments, the cellular constituent abundance data is gene expression data. In some embodiments, gene expression in a respective cell in the plurality of cells can be measured by sequencing the RNA transcripts of the cell and then counting the quantity of each gene transcript identified during the sequencing. In some embodiments, the gene transcripts sequenced and quantified include RNA, such as mRNA. In some embodiments, the gene transcripts sequenced and quantified include a downstream product of mRNA, such as a protein (e.g., a transcription factor). In general, as used herein, the term “gene transcript” may be used to denote any downstream product of gene transcription or translation, including post-translational modification, and “gene expression” and “cellular constituent abundance” may interchangeably be used to refer generally to any measure of gene transcripts.

[0138] In some embodiments, the abundance of a respective cellular constituent in the plurality of cellular constituents is RNA abundance (e.g., gene expression), and the abundance of the respective cellular constituent is determined by measuring polynucleotide levels of one or more nucleic acid molecules corresponding to the respective gene in the plurality of cells. The transcript levels of the respective gene can be determined from the amount of mRNA, or polynucleotides derived therefrom, present in the plurality of cells. Polynucleotides can be detected and quantitated by a variety of methods including, but not limited to, microarray analysis, polymerase chain reaction (PCR), reverse transcriptase polymerase chain reaction (RT-PCR), Northern blot, serial analysis of gene expression (SAGE), RNA switches, RNA fingerprinting, ligase chain reaction, Qbeta replicase, isothermal amplification method, strand displacement amplification, transcription based amplification systems, nuclease protection assays (Si nuclease or RNAse protection assays), and / or solid-state nanopore detection. See, e.g., Draghici, Data Analysis Tools for DNA Microarrays, Chapman and Hall / CRC, 2003; Simon et al., Design and Analysis of DNA Microarray Investigations, Springer, 2004; Real-Time PCR: Current Technology and Applications, Logan, Edwards, and Saunders eds., Caister Academic Press, 2009; Bustin A-Z of Quantitative PCR (IUL Biotechnology, No. 5), International University Line, 2004;Velculescu et al., (1995) Science 270: 484-487; Matsumura et al, (2005) Cell. Microbiol. 7: 11-18; Serial Analysis of Gene Expression (SAGE): Methods and Protocols (Methods in Molecular Biology), Humana Press, 2008; each of which is hereby incorporated herein by reference in its entirety.

[0139] In some embodiments, the abundance of each cellular constituent in the plurality of cellular constituents of the cellular constituent dataset 112 is obtained from expressed RNA or a nucleic acid derived therefrom (e.g., cDNA or amplified RNA derived from cDNA that incorporates an RNA polymerase promoter) from the plurality of cells, including naturally occurring nucleic acid molecules, as well as synthetic nucleic acid molecules. Thus, in some embodiments, the abundance of each cellular constituent in the plurality of cellular constituents of cellular constituent dataset is obtained from such non-limiting sources as total cellular RNA, poly(A)+ messenger RNA (mRNA) or a fraction thereof, cytoplasmic mRNA, or RNA transcribed from cDNA (e.g., cRNA). Methods for preparing total and poly(A)+ RNA are well known in the art, and are described generally, e.g., in Sambrook, et cd.. Molecular Cloning: A Laboratory Manual (3rd Edition, 2001). RNA can be extracted from a cell of interest using guanidinium thiocyanate lysis followed by CsCl centrifugation (see, e.g, Chirgwin et al., 1979, Biochemistry 18:5294-5299), a silica gel-based column (e.g, RNeasy (Qiagen, Valencia, Calif.) or StrataPrep (Stratagene, La Jolla, Calif.)), or using phenol and chloroform, as described in Ausubel etal., eds., 1989, Current Protocols In Molecular Biology, Vol. Ill, Green Publishing Associates, Inc., John Wiley & Sons, Inc., New York, at pp. 13.12.1-13.12.5). Poly(A)+ RNA can be selected, e.g, by selection with oligo-dT cellulose or, alternatively, by oligo-dT primed reverse transcription of total cellular RNA. RNA can be fragmented by methods known in the art, e.g., by incubation with ZnC12, to generate fragments of RNA.

[0140] In some embodiments, the abundance of each cellular constituent in the plurality of cellular constituents of the cellular constituent dataset 112 is determined by sequencing transcripts from the plurality of cells. In some embodiments, the abundance of each cellular constituent in the plurality of cellular constituents of the cellular constituent dataset 112 is determined by single-cell ribonucleic acid (RNA) sequencing (scRNA-seq), scTag-seq, single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq), CyTOF / SCoP, E-MS / Abseq, miRNA-seq, CITE-seq, and any combination thereof.

[0141] The cellular constituent abundance measurement technique can be selected based on the desired cellular constituents to be measured. For instance, scRNA-seq, scTag-seq, and miRNA-seq can be used to measure RNA expression. Specifically, scRNA-seq measures expression of RNA transcripts, scTag-seq allows detection of rare mRNA species, and miRNA-seq measures expression of micro-RNAs. CyTOF / SCoP and E-MS / Abseq can be used to measure protein expression in the cell. CITE-seq simultaneously measures both gene expression and protein expression in the cell, and scATAC-seq measures chromatinconformation in the cell. Table 1 below provides example protocols for performing each of the cellular constituent abundance measurement techniques described above.

[0142] Table 1 - Example Measurement Protocols

[0143] Block 226. Referring to block 226, in some embodiments, the plurality of cellular constituents consists of between 100 and 35,000 cellular constituents. In some embodiments, the plurality of cellular constituents comprises at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, atleast 6000, at least 7000, at least 8000, at least 9000, at least 10,000, at least 20,000, at least 30,000, at least 50,000, or more than 50,000 cellular constituents. In some embodiments, the plurality of cellular constituents comprises no more than 70,000, no more than 50,000, no more than 30,000, no more than 10,000, no more than 5000, no more than 1000, no more than 500, no more than 200, no more than 100, no more than 90, no more than 80, no more than 70, no more than 60, no more than 50, or no more than 40 cellular constituents. In some embodiments, the plurality of cellular constituents consists of between twenty and 10,000 cellular constituents. In some embodiments, the plurality of cellular constituents consists of between 100 and 35,000 cellular constituents. In some embodiments, the plurality of cellular constituents comprises from 5 to 20, from 20 to 50, from 50 to 100, from 100 to 200, from 200 to 500, from 500 to 1000, from 1000 to 5000, from 5000 to 10,000, or from 10,000 to 50,000 cellular constituents. In some embodiments, the plurality of cellular constituents falls within another range starting no lower than 5 cellular constituents and ending no higher than 70,000 cellular constituents.

[0144] As an example, in some embodiments, the plurality of cellular constituents comprises a plurality of genes, optionally measured at the RNA level. In some embodiments, the plurality of genes comprises at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, or at least 1000 genes. In some embodiments, the plurality of genes comprises at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, at least 10,000, at least 30,000, at least 50,000, or more than 50,000 genes. In some embodiments, the plurality of genes comprises from 5 to 20, from 20 to 50, from 50 to 100, from 100 to 200, from 200 to 500, from 500 to 1000, from 1000 to 5000, from 5000 to 10,000, or from 10,000 to 50,000 genes.

[0145] As another example, in some embodiments, the plurality of cellular constituents comprises a plurality of proteins. In some embodiments, the plurality of proteins comprises at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, or at least 1000 proteins. In some embodiments, the plurality of proteins comprises at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, at least 10,000, at least 30,000, at least 50,000, or more than 50,000 proteins. In some embodiments, the plurality of proteins comprises from 5 to 20, from 20 to 50, from 50 to 100, from 100 to 200, from 200 to 500, from 500 to 1000, from 1000 to 5000, from 5000 to 10,000, or from 10,000 to 50,000 proteins.

[0146] In some embodiments, the cellular constituent dataset 112 includes cellular constituent abundance values 122 for multiple instances of the plurality of cellular constituents. For example, in some embodiments the cellular constituent dataset 112 includes a separate set of cellular constituent abundance values 122 for the plurality of cellular constituents for each cell in the plurality of cells.

[0147] In some embodiments, the cellular constituent dataset 112 includes a separate set of cellular constituent abundance values 122 for the plurality of cellular constituents for each subset of cells in the plurality of cells that have the same concentration 116 / exposure time pair 118. In some such embodiments, for each respective cellular constituent, the measured cellular constituent abundance value for the respective cellular constituent in each cell in the respective subset of cells that have the same concentration 116 / exposure time pair 118 is averaged together to form the cellular constituent abundance value for the respective cellular constituent.

[0148] In some embodiments, the cellular constituent dataset 112 includes a separate set of cellular constituent abundance values 122 for the plurality of cellular constituents for each subset of cells in the plurality of cells that have the same concentration 116 / exposure time pair 118 from the same replica (e.g., the same well and / or aliquot of cells). In some such embodiments, for each respective cellular constituent, the measured cellular constituent abundance value for the respective cellular constituent in each cell in the respective subset of cells that have the same concentration 116 / exposure time pair 118 from the same replica is averaged together to form the cellular constituent abundance value for the respective cellular constituent. Thus, if there are three replicas, this will result in three distinct measurements of the same cellular constituent for a given concentration 116 / exposure time pair 118.

[0149] Block 228. Referring to block 228, in some embodiments, the cellular constituent dataset 112 is used to determine a respective pathway value for each pathway module 142 in a plurality of pathway modules, thereby obtaining a plurality of pathway values.

[0150] Each respective pathway module 142 in the plurality of pathway modules comprises an independent subset of the plurality of cellular constituents. For instance, referring to Figure IB, the independent subset of the plurality of cellular constituents for the pathway module 142-1 consists of cellular constituents 144-1-1, 144-1-2, ..., 144-1-P, whereas the independent subset of the plurality of cellular constituents for the pathway module 142-2 consists of cellular constituents 144-2-1, 144-2-2, ..., 144-1-W, where W and P are the same or different positive integers. As such, each independent subset of the plurality of cellular constituents for each respective pathway module comprises two or morecellular constituents. Moreover, each cellular constituent in the plurality of cellular constituents (of cellular constituent dataset 112) is in at least one independent subset of the plurality of cellular constituents associated with a pathway module 142 in the plurality of pathway modules. In some embodiments, a cellular constituent in the plurality of cellular constituents is in 2, 3, 4, 5, 6, 7, 9, 10, 11, 12, 13, 14, or 15 or more of the pathway modules in the plurality of pathway modules.

[0151] In some embodiments, the plurality of cellular constituents in the plurality of cellular constituents comprises 10 or more cellular constituents. Block 226 above provides additional suitable numbers of cellular constituents in the plurality of cellular constituents in accordance with various embodiments of the present disclosure.

[0152] In some embodiments, the plurality of cellular constituent abundance values comprises 5000 or more cellular constituent abundance values. In some embodiments, the plurality of cellular constituent abundance values comprises 10 or more, 50 or more, 100 or more, 250 or more, 500 or more, 1000 or more, 2000 or more, 5000 or more, 10,000 or more or 15,000 or more cellular constituent abundance values. In some such embodiments each of the cellular constituents is a different gene.

[0153] There is no requirement that each cellular constituent in a pathway module be unique. For instance, consider the case in which pathway module A contains cellular constituents, 1, 3 and 10. Other pathway module in the plurality of pathway module can contain these cellular constituents as well. Here, the term “independent” means that the subset of the plurality of cellular constituents in a particular pathway module is unique as a whole. Thus, considering the example pathway module A above, another pathway module in the plurality of pathway modules may contain cellular constituents, 1, 3 and 10 provided that it further contains other cellular constituents that pathway module A does not contain.Further considering the example pathway module A above, another pathway module in the plurality of pathway module modules may have cellular constituents, 1, 3 and 10 provided that it further contains other cellular constituents that pathway module A does not contain.

[0154] In some embodiments, each pathway module in the plurality of pathway module comprises the same or a different number of cellular constituents. In some embodiments, each respective independent subset of cellular constituents corresponding to each respective pathway module is a unique subset of cellular constituents (e.g., non-overlapping, where each cellular constituent in the plurality of cellular constituents is grouped into no more than one module). In some embodiments, a first pathway module has a first subset of cellular constituents that overlaps a second subset of cellular constituents corresponding to a secondpathway module (e.g., overlapping, where at least one cellular constituent in the plurality of cellular constituents is common to two or more different pathway modules).

[0155] In some embodiments, one or more of the pathway modules 142 have Wikipathway reference numbers. See Martens etal., 2021, “WikiPathways: connecting communities,” Nucleic Acids Research 8:49(D1) D613-D621, and Pico etal., 2008, “WikiPathways: Pathway Editing for the People,” PLoS Biol, each of which is hereby incorporated by reference. Such Wikipathway reference numbers are illustrated in Figures 8A and 8B. For example, in Figure 8A, one of the pathway modules 142 is the “Metapathway biotransformation Phase I and II” biological pathway and has a Wikipathway reference number of WP702. The independent subset of the plurality of cellular constituents that constitute the WP702 pathway is found at: https: / / www.wikipathways.org / pathways / WP702.htm1. They are: GSTA5, CYP4B 1, NDST4, GSTA4, AKR7A2 NAT12, NAT1, SULT1C2, UGT1 A5, NAT2, UGT1 A4, CYP39A1, CYP27A1, CYP2D6, NDST2, CYP2U1, CYP450, FM05, NDST3, HS6ST1, GSTT2, UGT1A7, CYP4F3, AKR1C2, CYP4X1, HS3ST1, MGST2, HS3ST3A1, CYP2A7, CYP4F22, GPX1, CHST7, TPMT, GLYAT, HNMT, BAAT, FM03, CYP2B6, CYP2C19, CHST1, GSTA1, CYP24A1, CHST3, UGT1A10, CYP4F12, UGT2B4, SULT1B1, UGT1A9, CYP27C1, CYP2A6, SULT1C4, CYP46A1, CYP2C9, CYP4V2, NAT6, GSTM4, GPX2, CYP11B1, CYP2C18, CYP4F11, GLYATL1, UGT1A1, CYP1A2, AKR1B1, HS3ST2, NAT14, GPX3, CYP1B1, GLYATL2, GSTO1, CYP20A1, CYP2S1, SULT1E1, CYP19A1, CYP21A2, AKR7A3, CYP26C1, SULT2A1, HS6ST2, CYP27B1, CYP26B1, GSTT1, AKR1A1, FM01, UGT2B7, NAT5, UGT2A2, INMT, NNMT, GSTM2, AKR1C3, CYP2W1, NAT10, CHST11, CHST9, SULT1C3, GAL3ST4, CYP3A7, UGT2B15, UGT2B11, CYP7A1, GSTZ1, CYP4F2, CHST2, CYP2F1, GAL3ST1, HS3ST5, GSTK1, GAL3ST3, HS2ST1, CYP8B1, CYP1A1, GSR, SULT4A1, HS6ST3, CYP11B2, GSTM5, CYP4Z1, AKR1C4, UGT2A1, NDST1, AKR1C1, KCNAB1, GSTCD, CHST13, CYP7B1, CYP11A1, CYP2C8, UGT2A3, CHST10, SULT1C1, EPHX1, NAT9, NAT8L, SULT1A2, NAT13, HS3ST4, UGT1A6, GPX4, CYP2R1, HS3ST3B1, SULT2B1, GSS, GPX5, Glutathione transferase, CYP4F8, EPHX2, CHST5, UGT2B28, CYP3A4, GSTM1, KCNAB3, FMO2, GSTA3, KCNAB2, NATH, FMO4, CYP26A1, SULT1A1, CYP2E1, CHST14, SULT6B1, HS3ST6, CYP17A1, CHST4, MGST1, UGT2B17, SULT1A4, KR1B10, NAT8, CYP3A43, SULT1A3, UGT1A8, CYP3A5, GSTA2, GSTP1, COMT, CHST6, UGT1A3, CYP2J2, MGST3, AKR1D1, GSTO2, GSTM3, GAL3ST2, GSTT2B, CHST8, CYP2A13, CYP51A1, and CHST12.

[0156] In another example, in Figure 8A, one of the pathway modules 142 is the “Tamoxifen metabolism” biological pathway and has a Wikipathway reference number of WP691. The independent subset of the plurality of cellular constituents that constitute the WP691 pathway is found at: https: / / www.wikipathways.org / pathways / WP691.html. They are: CYP2D6, CYP2A6, SULT2A1, CYP2C9, CYP3A5, CYP2C19, CYP3A4, CYP3A4, UGT1A10, CYP3A4, SULT2A1, SULT2A1, CYP1A2, CYP3A4, UGT2B7, CYP2D6, SULT1E1, SULT1A1, FM01, CYP2C9, CYP2E1, SULT1A1, FM03, SULT1E1, CYP3A4, UGT1A4, CYP3A5, CYP2C19, UGT1A4, CYP1B1, UGT1A4, UGT1A8, CYP2C19, CYP3A4, CYP2D6, CYP1A1, CYP2D6, CYP2C8, UGT2B15, and CYP3A4. In addition to these genes, the WP691 pathway has a number of metabolites, such as amoxifen-N-oxide, ci s-4-hydroxytam oxifen, and alpha-hydroxy-N-desmethyltamoxifen. In some embodiments where the Wikipathway includes both genes and metabolites, the independent subset of cellular constituents for the corresponding pathway module 142 includes both the metabolites and the genes. In some embodiments, the independent subset of cellular constituents for the corresponding pathway module 142 includes only the genes. In some embodiments, the corresponding pathway module 142 for a Wikipathway consists of only a subset of the genes in a corresponding Wikipathway, such as five or more, ten or more or between 3 and 30 of the corresponding Wikipathway. In some embodiments, the corresponding pathway module 142 for a Wikipathway consists of only those genes in the corresponding Wikipathway that differentially express across a plurality of annotated cell states. Such annotated cells states and differentially expression is described in block 236 below.

[0157] In some embodiments, the plurality of pathway modules comprises 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 or more of the Wikipathways listed in Figures 8 A and 8B.

[0158] In some embodiments, one or more of the pathway modules 142 are KEGG pathways. For instance, in some embodiments the subset of cellular constituents of such pathway modules consists of the cellular constituent in corresponding KEGG pathways. In some embodiments, one or more of the pathway modules 142 are human KEGG pathways. Such pathways are disclosed at

[0159] In some embodiments, one or more of the pathway modules 142 are Reactome pathways. For instance, in some embodiments the subset of cellular constituents of such pathway modules consists of the cellular constituent in corresponding Reactome pathways. In some embodiments, one or more of the pathway modules 142 are human Reactome pathways. In some embodiments, the corresponding pathway module 142 for a Reactomepathway consists of only a subset of the genes in the corresponding Reactome pathway, such as five or more, ten or more or between 3 and 30 of the Reactome pathway. In some embodiments, the corresponding pathway module 142 for a Reactome pathway consists of only those genes in the corresponding Reactome pathway that differentially express across the plurality of annotated cell states. Such differential expression and annotated cells states is described in block 236 below. Reactome pathways are disclosed at htps: / / rea ome.org .See also Gillespie et al., 20>22, “The reactome pathway knowledgebase 2022”, Nucleic Acids Research 50(Dl):D687-D692, which is hereby incorporated by reference.

[0160] In some embodiments, one or more of the pathway modules 142 are Pathway Commons pathways. For instance, in some embodiments the subset of cellular constituents of such pathway modules consists of the cellular constituent in corresponding Pathway Commons pathways. In some embodiments, one or more of the pathway modules 142 are human Pathway Commons pathways. In some embodiments, the corresponding pathway module 142 for a Pathway Commons pathway consists of only a subset of the genes in the corresponding Pathway Commons pathway, such as five or more, ten or more or between 3 and 30 of the Pathway Commons pathway. In some embodiments, the corresponding pathway module 142 for a Pathway Commons pathway consists of only those genes in the corresponding Pathway Commons that differentially express across a plurality of annotated cell states. Such differential expression and annotated cells states is described in block 236 below. Pathway Commons pathways are disclosed at ht s: / / www. athwaycommoHS.or . See also Rodchenkov etal., 2020, “Pathway Commons 2019 Update: integration, analysis and exploration of pathway data,” Nucleic Acids Research, Volume 48, Issue DI, 08 January 2020, Pages D489-D497, which is hereby incorporated by reference.

[0161] Block 230. Referring to block 230, in some embodiments, each independent subset of the plurality of cellular constituents for each respective pathway module comprises between two cellular constituents and three hundred cellular constituents. In some embodiments, the independent subset of the plurality of cellular constituents in a respective pathway module comprises three, four, or five or more cellular constituents. In some embodiments, the independent subset of the plurality of cellular constituents in a respective pathway module in the plurality of pathway modules comprises at least 2, at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, or at least 500 cellular constituents. In some embodiments, the independent subset of the plurality of cellular constituents comprises no more than no more than 500, no more than 200, no morethan 100, no more than 90, no more than 80, no more than 70, no more than 60, or no more than 50 cellular constituents. In some embodiments, the independent subset of the plurality of cellular constituents comprises from 5 to 100, from 2 to 300, from 20 to 500, or from 200 to 1000 cellular constituents. In some embodiments, the independent subset of the plurality of cellular constituents falls within another range starting no lower than 2 cellular constituents and ending no higher than 2000 cellular constituents.

[0162] In some embodiments, the independent subset of the plurality of cellular constituents in the respective pathway module consists of cellular constituents in a cellular process (e.g., a molecular pathway). For instance, in some embodiments, the independent subset of the plurality of cellular constituents in a respective pathway module consists of between two and 20 cellular constituents in a molecular pathway associated with DILI. Figures 8A and 8B name some of these molecular pathways in accordance with some such embodiments.

[0163] Referring to block 232, in some embodiments, each respective pathway value of each respective pathway module 142 is a measure of central tendency of an abundance of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents of the respective pathway module 142, across the cellular constituent dataset. As used here, the term “across the cellular constituent dataset” means the measured abundances for the independent subset of cellular constituents in the cells exposed to the compound at a particular concentration 116 for a particular exposure time 118. If the cellular constituent dataset includes data for the same exposure time at four different concentrations, than in such embodiments there would be four different pathway values for respective pathway module 142, one for each of the four different exposure times. If the cellular constituent dataset includes data for four different exposure times each at two different concentrations, than in such embodiments there would be eight different pathway values for the respective pathway module 142, one for each of the eight unique combinations of exposure times and concentrations.

[0164] In some embodiments the measure of central tendency is a mean, median, mode, a weighted mean, weighted median, weighted mode of the measured abundance of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents of the respective pathway module. For instance, in the case where there are five cellular constituents in the independent subset of cellular constituents for a given pathway module, a measure of central tendency of the measured cellular constituent abundance for the five cellular constituents (at a given concentration 116and exposure time 118) is taken as the pathway value for the respective pathway module (at the given concentration 116 and exposure time 118) in such embodiments.

[0165] Referring to block 234, in alternative embodiments, each respective pathway value of each respective pathway module is based on a p-value. In some such embodiments the plurality of cellular constituents are genes. Moreover, in some embodiments where the plurality of cellular constituents are genes and pathway enrichment analysis is used to obtain the respective pathway values for each pathway module in the plurality of pathway modules. The pathway enrichment analysis is a statistical method that determines whether a particular pathway module is overrepresented or enriched with differentially expressed genes (DEGs) in RNA-seq data. In such embodiments, the pathway value of a respective pathway module is determined by first determining a p-value for the respective pathway module by pathway enrichment analysis using a statistical test. This statistical test determines the significance of the observed enrichment of DEGs within the respective pathway module. In some embodiments, the hypergeometric test is used to compute the p-value for pathway module. The hypergeometric test determines the probability of obtaining the observed number of DEGs in a specific pathway module by chance, given the total number of genes in the pathway and the total number of DEGs in the entire dataset. To transform the p-value of a pathway module into a pathway value for the pathway module, in some embodiments, a negative logarithm of the adjusted p-value, or False Discovery Rate (FDR), is calculated and used as the pathway value. The FDR corrects for multiple testing in pathway enrichment analysis to control the proportion of false positives. Accordingly, in more detail, in some embodiments where the hypergeometric test is used, the following parameters are defined:• N: the number of genes in the plurality of cellular constituents (e.g, approximately thirty thousand in some embodiments),• n: the number of genes annotated to the particular pathway module (e.g., in some embodiments the particular pathway module contains the genes of a single pathway of interest),• K: the total number of DEGs in the plurality of cellular constituents (e.g., in the genome / dataset), and• k: Total number of DEGs annotated to the particular pathway module t.

[0166] In some embodiments, the probability of observing exactly k DEGs in the pathway module out of n total genes in the pathway module and K DEGs out of N total genes in the dataset is calculated using the hypergeometric probability formula.

[0167] In some embodiments, a multiple testing correction is applied to a p-value of a pathway module. In some embodiments the multiple testing correction is a Benjamini- Hochberg procedure. The multiple testing correction adjusts the p-values for multiplecomparisons. The result is a set of adjusted p-values or FDR (false discovery rate) values, each adjusted p-value in the set of adjusted p-values representing the significance of a corresponding pathway module in the plurality of pathway modules while considering the number of pathway modules tested.

[0168] In some embodiments, to transform the FDR values into pathway values, a negative logarithm (base 10) of each FDR value is taken. In some embodiments, this transformation helps emphasize significant enrichments while linearizing the data. In some embodiments the negative logarithm(base 10) is computed using the formula -loglO(FDR). The resulting -loglO(FDR) values represent the pathway values for the plurality of pathway modules. Higher -loglO(FDR) value indicate more significant enrichment of DEGs in a pathway module.

[0169] Referring to block 236, in some embodiments, the method further comprises identifying the plurality of pathway modules. That is, in some embodiments of the present disclosure, a pathway module consists of cellular constituents that is not documented in a pathway in a publicly curated database such as Wikipathway, KEGG, Reactome, or Pathway Commons.

[0170] In some embodiments, a pathway module consists of at least 20, at least 30, at least 40, at least 50 or all 64 of the genes: RNF187, CHAC1, CARS, GPD1L, AARS, CSTA, TRIB3, LONP1, RHBDD1, ANKRD30BL, FGF21, PIR, FXYD5, ARL6IP6, ASNS, PHGDH, SLC31A1, SLC7A1, NARS, MTHFD2, SARS, JAK1, XPOT, PSAT1, EIF4EBP1, MOCOS, GARS, SLC7A11, SESN2, RHOF, SLFN5, PCK2, MKX, AGPAT9, LCN2, STC2, MKNK2, TOM1L1, DDIT4, LOC100128822, RSL24D1, YARS, M6PR, CTH, ANO10, C19orf80, ECU, FBXO25, NCF2, FGF2, TMEM206, PRNP, TUBE1, MARS, EFHD1, LRRC8B, PPM1E, PVR, SLC43A1, EMP1, SLC7A5, MEI, UNC5B, and ATF5. This set of genes does not correspond to any of documented pathways in Wikipathway, KEGG, Reactome, or Pathway Commons.

[0171] In some embodiments, a pathway module is identified by obtaining one or more first datasets (e.g., in electronic form). The one or more first datasets comprise, for each respective primary human hepatocyte cell in a plurality of primary human hepatocyte cells, for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell. In some such embodiments, the plurality of primary human hepatocyte cells comprises 20, 30, 40, 50, 100, or 100 or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states. From this data theplurality of cellular constituents is identified from the one or more first datasets as a plurality of differentially expressed cellular constituents across the one or more first datasets. Differentially expressed cellular constituents are those that express differently across the plurality of annotated cell states. Differentially expressed cellular constituents are sought because they are more informative for determining DILI toxicity then cellular constituents whose expression is invariant across the plurality of annotated cell states.

[0172] Several metrics and statistical methods can be employed to assess the differential expression of cellular constituents.

[0173] One such metric is fold change. Fold change measures the ratio of expression levels of a cellular constituent between two conditions. It is calculated as the ratio of the expression level of the cellular constituent in one condition to the expression level in another condition. A fold change of 1 indicates no differential expression, whereas values greater than 1 or less than 1 indicate upregulation or downregulation, respectively. In instances where more than two conditions are tested, a measure of central tendency of the fold change of every unique pair of conditions for a given cellular constituent can be used to determine the differential expression of the given cellular constituent. The cellular constituents can then be ranked in terms of the absolute value of these fold changes and the top N number of cellular constituents, where N is a positive integer such as 50, 100, 500, or a number between 10 and 2000, in terms of absolute fold change can be deemed the cellular constituents that are differentially expressed.

[0174] Another such metric is log fold change. Log Fold change measures is the logarithm of the ratio of expression levels of a cellular constituent between two conditions. It is calculated as the ratio of the expression level of the cellular constituent in one condition to the expression level in another condition. In instances where more than two conditions are tested, a measure of central tendency of the log fold change of every unique pair of conditions for a given cellular constituent can be used to determine the differential expression of the given cellular constituent. The cellular constituents can then be ranked in terms of the absolute value of these log fold changes and the top N number of cellular constituents, where N is a positive integer such as 50, 100, 500, or a number between 10 and 2000, in terms of absolute log fold change, can be deemed the cellular constituents that are differentially expressed.

[0175] Another such metric is p-value. The p-value is a statistical measure that quantifies the probability of observing the observed differential cellular constituent expression across the plurality of annotated cells states by chance alone. It can, for example, be calculatedusing statistical tests such as t-tests, ANOVA, or non-parametric tests. A lower p-value indicates a higher level of statistical significance and suggests that the observed differential expression across the plurality of annotated cells states is unlikely to be due to random chance. The cellular constituents can then be ranked in terms of p-value and the top N number of cellular constituents (indicating the N cellular constituents with the smallest p- values), where N is a positive integer such as 50, 100, 500, or a number between 10 and 2000, can be deemed the cellular constituents that are differentially expressed. In some embodiments, rather than a p-value, an adjusted p-value is used. For example, in some embodiments a Benjamini -Hochberg p-value is used to control the false discovery rates. In other embodiments a Bonferroni corrected p-value is used to control family-wise error rate (FWER).

[0176] Once the differentially expressed cellular constituents are identified they can be assigned to pathway modules. In some such embodiments, the differentially expressed cellular constituents are compared to the sets of cellular constituents associated with pathways in a publicly curated database such as Wikipathway, KEGG, Reactome, or Pathway Commons. When a subset of the differentially expressed cellular constituents (e.g., at least 3, 4, 5, 6, 7, 8, 9, or 10 differentially expressed cellular constituents) is found to all be present in such a pathway, that pathway is selected as one of the pathway module 142.

[0177] In some embodiments each of the pathway modules represents a cluster of differentially expressed cellular constituents. In some embodiments such clusters are identified by obtaining one or more first datasets in electronic form, the one or more first datasets comprising: for each respective primary human hepatocyte cell in a plurality of primary human hepatocyte cells, where the plurality of primary human hepatocyte cells comprises twenty or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states: for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell. In this way a plurality of vectors are accessed or formed. Each respective vector in the plurality of vectors (i) corresponding to a respective cellular constituent in the plurality of (differentially expressed) cellular constituents and (ii) comprising a corresponding plurality of elements, each respective element in the corresponding plurality of elements representing an abundance of the respective cellular constituent in a corresponding primary human hepatocyte cell in the plurality of primary human hepatocyte cells. The plurality of vectors is clustered thereby forming a plurality of clusters. Each cluster in the plurality of clusters consists of an independent subset of theplurality of cellular constituents of a respective cellular constituent module in the plurality of cellular constituent modules.

[0178] In some embodiments each of the pathway modules represents a cluster of differentially expressed cellular constituents. In some embodiments such clusters are identified by obtaining one or more first datasets in electronic form, the one or more first datasets comprising: for each respective primary human hepatocyte cell in a plurality of primary human hepatocyte cells, where the plurality of primary human hepatocyte cells comprises twenty or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states: for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell. From this a consensus cellular constituent module matrix is generated using the one or more first datasets using consensus non-negative matrix factorization. The independent subset of the plurality of cellular constituents in each respective cellular constituent module in the plurality of cellular constituent modules is then identified from the consensus matrix using multiple least squares regression of the corresponding abundances of the cellular constituents in the one or more first datasets against the consensus GEP usage matrix.

[0179] In some such embodiments, each independent subset of cellular constituents in the plurality of cellular constituents is associated with a corresponding pathway for a pathway module in the plurality of pathway modules. For example, in one embodiment an independent subset of cellular constituents identified using the consensus matrix approach above is ARSE, ARG1, ANKS4B, ABHD2, KANK1, TACSTD2, SALL1, LPIN2, SEC16B, CEBPD, NAV2, ERBB3, Clorfl l5, PLA2G12B, ABCB4, CYP39A1, TAT, METTL7A, RPRD1B, PPAP2A, CYP8B1, MTCP1NB, DEPDC7, DUSP16, RCL1, ANKRD35, KLHDC2, ABCG2, CIDEC, Cl lorf54, Cl lorf75, RAB17, PEBP1, PET117, AKAP12, TBX15, GALM, LEAP2, CROCCP2, RNASE4, IRF2, SORBS1, TMEM37, CYP27A1, CLRN3, RAPGEF5, AQP3, C2orf72, PCCA, HSD17B11, NSMCE2, RHOB, FAM134B, AGTR1, AADAC, HMGB3, PANK1, and HPGD. Analysis of this subset of cellular constituents indicates that some of the cellular constituents are found in the KEGG primary bile acid biosynthesis pathway. Thus, the pathway module for this subset of cellular constituents is associated with the KEGG primary bile acid biosynthesis pathway. The import of such assignments is to identify a biological basis for the clusters of cellular constituents that are assigned to pathway modules in such embodiments. However, in embodiments where the independent subsets of cellular constituents are identified throughclustering, there is no requirement that such subsets be associated with a known molecular pathway. In fact, such clusters may represent molecular mechanism associated with DILI that have not yet been discovered.

[0180] Referring to block 238, in some embodiments, an annotated cell state in the plurality of annotated cell states is an exposure of a cell in the plurality of primary human hepatocyte cells to a training compound under an exposure condition (e.g., the exposure condition is a duration of exposure, a concentration of a training compound, or a combination of a duration of exposure and a concentration of the training compound). The examples section below describes the exposure of cells to 120 different such training compounds.

[0181] Referring to block 240, in some embodiments, the plurality of annotated cell states includes exposure to a first training compound at a plurality of different concentrations e.g., the plurality of different concentrations range from (i) a Cmax or an imputed Cmax of the first training compound to (ii) an ICio of the first training compound). In some embodiments, the plurality of different concentrations is 2, 3, 4, 5, 6, 7, 8, 9, or 10 different concentrations ranging from 0.1 pM to 1 pM. In some embodiments, the plurality of different concentrations is 2, 3, 4, 5, 6, 7, 8, 9, or 10 different concentrations ranging from 0.01 pM to 1 pM. In some embodiments, the plurality of different concentrations is 2, 3, 4, 5, 6, 7, 8, 9, or 10 different concentrations ranging from 0.1 pM to 10 pM. In some embodiments, the lowest concentration of the plurality of different concentrations is 0.001 pM, 0.01 pM, 0.01 pM, or 1 pM. In some embodiments, the highest concentration used in the plurality of different concentrations is 100 pM, 50 pM, 10 pM, 1 pM, 0.5 pM, or 0.1 pM.

[0182] Block 242. Referring to block 242, in some embodiments, the plurality of annotated cell states includes exposure to a first training compound at a plurality of different durations of exposure. In some embodiments, the plurality of annotated cell states includes exposure to a plurality of different training compound each at the same concentration and at a plurality of different durations of exposure. In some embodiments, the plurality of annotated cell states includes exposure to a plurality of different training compound at several different concentrations and several different durations of exposure.

[0183] In some embodiments, the plurality of annotated states comprises 2 or more, 3 or more, 4 or more, 5 or more, 6 or more, 7 or more, 8 or more, 9 or more, 10 or more, 11 or more, 12 or more, 13 or more, 14 or more, 15 or more, 16 or more, 17 or more, 18 or more, 19 or more, 20 or more, 30 or more, 40 or more, 50 or more, 60 or more, 70 or more, 80 or more, 100 or more, 200 or more, or 1000 or more annotated cell states.

[0184] Referring to block 244, in some embodiments, the plurality of annotated cell states includes exposure to each respective training compound in a plurality of training compounds at a corresponding plurality of different concentrations (e.g., the corresponding plurality of different concentrations range from (i) a Cmax of the respective training compound to (ii) an ICio of the respective training compound).

[0185] In some embodiments an annotated cell state is an exposure of a cell to a compound under an exposure condition (e.g., a duration of exposure, a concentration of the compound, or a combination of a duration of exposure and a concentration of the compound).

[0186] Referring to block 246, in some embodiments, the plurality of annotated cell states includes exposure to each training compound in the plurality of training compounds at a plurality of different durations of exposure (e.g., 4 hours, 8 hours, 16 hours, 24 hours or some subset thereof).

[0187] Block 248. Referring to block 248, in some embodiments, the plurality of pathway modules consists of between 10 and 2000 pathway modules. In some embodiments, the plurality of pathway modules comprises five or more pathway modules. In some embodiments, the plurality of pathway modules comprises 20 or more, 30 or more, 40 or more, 50 or more, 60 or more, 70 or more, 80 or more 90 or more, or 100 or more pathway modules.

[0188] In some embodiments, the plurality of pathway modules comprises at least 5, at least 10, at least 15, at least 20, at least 25, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, at least 1000, at least 2000, or at least 3000 pathway modules. In some embodiments, the plurality of pathway modules comprises no more than 2000, no more than 1000, no more than 500, no more than 300, no more than 200, no more than 100, no more than 90, no more than 80, no more than 70, no more than 60, or no more than 50 pathway modules. In some embodiments, the plurality of pathway modules consists of between 50 and 500 pathway modules. In some embodiments, the plurality of pathway modules comprises from 5 to 20, from 20 to 50, from 50 to 100, from 100 to 200, from 200 to 500, from 500 to 1000, or from 1000 to 5000 pathway modules. In some embodiments, the plurality of cellular constituent modules falls within another range starting no lower than 5 cellular constituent modules and ending no higher than 10,000 pathway modules.

[0189] Block 249. Referring to block 249, in some embodiments, the plurality of pathway values is inputted into each first model in one or more first model 130 therebyretrieving, as first model output 134 from each of the one or more first models, one or more first values. That is, each first model outputs a first value and the pathway values calculated for each of the pathway modules 142 of the plurality of pathway modules is concurrently inputted into each of the one or more first models.

[0190] In some embodiments each first model in the one or more first models is a logistic regression model that is trained using a training dataset of compounds for which the DILI risk of each compound in the training dataset is known. In some such embodiments the logistic regression model includes a different weight for each one of the pathway module values.

[0191] For instance, the Examples section below describes training the first logistic regression model on gene expression data and toxicological information for 120 compounds using data from the Japanese Toxicogenomics Project (TGP), which includes in vitro gene expression data, toxicological information and pathological data for these compounds. This data is available on the Internet at

[0192] In some embodiments, training each of the one or more first models 130 further uses a regularization on each model parameter. For example, in some embodiments, a regularization is performed by adding a penalty to the loss function, where the penalty is proportional to the values of the parameters in the first model. Such practice can result in a more generalized model and reduce overfitting of the data. In some embodiments, the regularization includes an LI or L2 penalty. For example, in some preferred embodiments, the regularization includes an L2 penalty on lower and upper parameters. In some embodiments, the regularization comprises spatial regularization (e.g., determined based on a priori and / or experimental knowledge) or dropout regularization. In some embodiments, the regularization comprises penalties that are independently optimized.

[0193] Generally, training each of the one or more first models comprises updating the plurality of parameters (e.g., weights 132-1, 132-2, . . ., 132-K, where K is a positive integer) for a given first model 130 in response to such training data. In some embodiments this is done through backpropagation (e.g., gradient descent). First, a forward propagation is performed, in which input data (the plurality of pathway values for each respective training compound in a plurality of training compounds) is accepted one by one into the first model 130, and the first model output 134 for each of these training compounds is calculated based on the initial set of model parameters (e.g., weights 132). In some embodiments, prior to training, the model parameters (e.g., weights 132 and / or hyperparameters) are randomly assigned (e.g., initialized) for a given first model. In some embodiments, values for theseparameters are transferred from a previously saved plurality of parameters or from a pretrained model (e.g., by transfer learning).

[0194] In some embodiments, a backward pass is then performed by calculating an error gradient for a given first model, where the error is determined by calculating a loss (e.g., error) based on the given first model output 134 (e.g., the predicted DILI risk) and the actual known DILI risk for each training compound. Parameters (e.g., weights) are then updated by adjusting the values of the parameters based on the calculated loss, thereby training the given first model.

[0195] For example, in some general embodiments of machine learning, backpropagation is a method of training a plurality of weights of a model. The output of the untrained model (e.g., the predicted DILI risk) is first generated using a set of arbitrarily selected initial weights upon inputting the plurality of pathway values. The output (predicted DILI risk) is then compared with the original input (e.g., the actual DILI risk) for each training compound by evaluating an error function to compute an error (e.g., using a loss function). The weights are then updated such that the error is minimized (e.g., according to the loss function). In some embodiments, any one of a variety of backpropagation algorithms and / or methods are used to update the plurality of weights, as will be apparent to one skilled in the art.

[0196] In some embodiments, the loss function is mean square error, quadratic loss, mean absolute error, mean bias error, hinge, multi-class support vector machine, and / or crossentropy. In some embodiments, training a given first model comprises computing an error in accordance with a gradient descent algorithm and / or a minimization function. In some embodiments, training a given first model comprises computing a plurality of errors using a plurality of loss functions. In some embodiments, each loss function in a plurality of loss functions receives a same or a different weighting factor.

[0197] In some embodiments, the error function is used to update one or more parameters (e.g., weights) in a given first model by adjusting the value of the one or more parameters by an amount proportional to the calculated loss, thereby training the given first model. In some embodiments, the amount by which the parameters are adjusted is metered by a learning rate hyperparameter that dictates the degree or severity to which parameters are updated (e.g., smaller or larger adjustments). Thus, in some embodiments, the training updates all or a subset of the plurality of parameters based on a learning rate. In some embodiments, the learning rate is a differential learning rate.

[0198] In some embodiments, the training process, including adjusting the plurality of parameters associated with a given first model (e.g., responsive to the difference between thepredicted and actual DILI risks), is repeated for each training instance in a plurality of training instances.

[0199] In some embodiments, the plurality of training instances comprises at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 50, at least 100, at least 500, at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, or at least 7500 training instances. In some embodiments, the plurality of training instances comprises no more than 10,000, no more than 5000, no more than 1000, no more than 500, no more than 100, or no more than 50 training instances. In some embodiments, the plurality of training instances comprises from 3 to 10, from 5 to 100, from 100 to 5000, or from 1000 to 10,000 training instances. In some embodiments, the plurality of training instances falls within another range starting no lower than 3 training instances and ending no higher than 10,000 training instances.

[0200] In some such embodiments, the training includes repeating the adjustment of the parameters 132 of a given first model 130 (e.g., via backpropagation) over a plurality of training instances, therefore increasing the model’s accuracy in determining a compounds DILI risk.

[0201] In some embodiments, the model training trains the given first model following a first evaluation of an error function. In some such embodiments, the model training trains the given first model after first updating one or more parameters based on a first evaluation of an error function. In some alternative embodiments, the model training trains the given first model through at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 100, at least 500, at least 1000, at least 10,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, or at least 1 million evaluations of an error function. In some such embodiments, the model training trains the given first model through at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 100, at least 500, at least 1000, at least 10,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, or at least 1 million updatings of one or more parameters based on the at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 100, at least 500, at least 1000, at least 10,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, or at least 1 million evaluations of an error function.

[0202] In some embodiments, the model training is repeated until the given first model satisfies a minimum performance requirement. For example, in some embodiments, themodel training is repeated until the error calculated for the trained model, following evaluation of the error function (e.g., a difference between a predicted and an actual DILI risk) satisfies an error threshold. In some embodiments, the error calculated by the error function satisfies the error threshold when the error is less than 20 percent, less than 18 percent, less than 15 percent, less than 10 percent, less than 5 percent, or less than 3 percent.

[0203] In an example embodiment, the given first model is trained using categorical cross-entropy loss in a multi-task formulation, in which each covariate of a plurality of covariates corresponds to a cost function in plurality of cost functions and each respective cost function in the plurality of cost functions has a common weighting factor.

[0204] Referring to block 250, in some embodiments, the given first model is a logistic regression model. Logistic regression models are disclosed in Agresti, An Introduction to Categorical Data Analysis, 1996, Chapter 5, pp. 103-144, John Wiley & Son, New York, which is hereby incorporated by reference.

[0205] Block 252. Referring to block 252, in some embodiments, each first model in the one or more first models is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model. Details of such models are disclosed above in the definitions section.

[0206] In some embodiments, each first model 130 in the one or more first models comprises a plurality (e.g., 10 or more, 20 or more, 30 or more, 50 or more, 100 or more, 200 or more, 300 or more, 500 or more, 1000 or more, or 10,000 or more) of parameters 132.

[0207] In some embodiments, each first model 130 in the one or more first models comprises a plurality of parameters 132 (e.g., weights and / or hyperparameters). In some embodiments, the plurality of parameters for each first model in the one or more first models comprises at least 10, at least 50, at least 100, at least 500, at least 1000, at least 2000, at least 5000, at least 10,000, at least 20,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, at least 1 million, at least 2 million, at least 3 million, at least 4 million or at least 5 million parameters. In some embodiments, the plurality of parameters for each first model in the one or more first models comprises no more than 8 million, no more than 5 million, no more than 4 million, no more than 1 million, no more than 500,000, no more than 100,000, no more than 50,000, no more than 10,000, no more than 5000, no more than 1000, or no more than 500 parameters. In some embodiments, the plurality of parameters for each first model in the one or more first models comprises from 10 to 5000, from 500 to 10,000,from 10,000 to 500,000, from 20,000 to 1 million, or from 1 million to 5 million parameters. In some embodiments, the plurality of parameters for each first model in the one or more first models falls within another range starting no lower than 10 parameters and ending no higher than 8 million parameters.

[0208] In some embodiments, the training of each first model in the one or more first models is further characterized by one or more hyperparameters (e.g., one or more values that may be tuned during training). In some embodiments, the hyperparameter values are tuned (e.g., adjusted) during training. In some embodiments, the hyperparameter values are determined based on the specific elements of a training dataset and / or one or more inputs (e.g., cells, cellular constituent modules, covariates, etc.). In some embodiments, the hyperparameter values are determined using experimental optimization. In some embodiments, the hyperparameter values are determined using a hyperparameter sweep. In some embodiments, the hyperparameter values are assigned based on prior template or default values.

[0209] In some embodiments, a respective hyperparameter of the one or more hyperparameters comprises a learning rate. In some embodiments, the learning rate is at least 0.0001, at least 0.0005, at least 0.001, at least 0.005, at least 0.01, at least 0.05, at least 0.1, at least 0.2, at least 0.3, at least 0.4, at least 0.5, at least 0.6, at least 0.7, at least 0.8, at least 0.9, or at least 1. In some embodiments, the learning rate is no more than 1, no more than 0.9, no more than 0.8, no more than 0.7, no more than 0.6, no more than 0.5, no more than 0.4, no more than 0.3, no more than 0.2, no more than 0.1 no more than 0.05, no more than 0.01, or less. In some embodiments, the learning rate is from 0.0001 to 0.01, from 0.001 to 0.5, from 0.001 to 0.01, from 0.005 to 0.8, or from 0.005 to 1. In some embodiments, the learning rate falls within another range starting no lower than 0.0001 and ending no higher than 1. In some embodiments, the one or more hyperparameters further include regularization strength (e.g., L2 weight penalty, dropout rate, etc.). For instance, in some embodiments, each first model in the one or more first models (e.g., a neural network) is trained using a regularization on a corresponding parameter (e.g., weight) of each hidden neuron in the plurality of hidden neurons. In some embodiments, the regularization includes an LI or L2 penalty.

[0210] In some embodiments, a respective hyperparameter of the one or more hyperparameters is the identity (type) of loss function. In some embodiments, the loss function is mean square error, flattened mean square error, quadratic loss, mean absolute error, mean bias error, hinge, multi-class support vector machine, and / or cross-entropy. Insome embodiments, the loss function is a gradient descent algorithm and / or a minimization function.

[0211] In some embodiments, each first model in the one or more first models is associated with one or more activation functions (e.g., in the case where the first model is a neural network). In some embodiments, an activation function in the one or more activation functions is tanh, sigmoid, softmax, Gaussian, Boltzmann-weighted averaging, absolute value, linear, rectified linear unit (ReLU), bounded rectified linear, soft rectified linear, parameterized rectified linear, average, max, min, sign, square, square root, multiquadric, inverse quadratic, inverse multiquadric, polyharmonic spline, swish, mish, Gaussian error linear unit (GeLU), and / or thin plate spline. In some embodiments, each first model in the one or more first models outputs a predicted DILI risk responsive to the inputting of the pathway values into the model.

[0212] Blocks 254 -260. Referring to block 254, in some embodiments, the plurality of cellular constituent abundance values 122 is inputted into each of one or more second models 136 (e.g., 136-1, 136-2, ..., 136-T, where T is a positive integer) thereby retrieving, as second model output from the one or more second models, one or more second values (e.g., 140-1, 140-2, . . ., 140-T). Referring to block 256, in some embodiments, the one or more second models is a single logistic regression model. Referring to block 258, in some embodiments, the one or more second models a plurality of logistic regression models. In some such embodiments the plurality of logistic regression models is between four and 100 logistic regression models. Referring to block 260, in some embodiments, a first subset of the plurality of logistic regression models in the plurality of second models is a plurality of LI regression models, and a second subset of the plurality of logistic regression models in the plurality of second models is a plurality of L2 regression models. In some such embodiments, the first subset of the plurality of logistic regression models is between two and fifty logistic regression models. In some such embodiments, the second subset of the plurality of logistic regression models is between two and fifty logistic regression models.

[0213] In some embodiments each second model 136 is a logistic regression model that is trained using a training dataset of compounds for which the DILI risk of each compound in the training dataset is known. For instance, the Examples section below describes training a plurality of second logistic regression models on gene expression data and toxicological information for 120 compounds using data from the Japanese Toxicogenomics Project (TGP), which includes gene expression data, toxicological information and pathological data forthese 120 compounds. This data is available on the Internet at http: / / camda2.Q2 l.bioinf.jku.a doku.php / tgp prepro.

[0214] In some embodiments, training each of the one or more second models 136 further uses a regularization on each model parameter 138 in each of the one or more second models. For example, in some embodiments, a regularization is performed by adding a penalty to the loss function, where the penalty is proportional to the values of the parameters 138 in the second model 136. Such practice can result in a more generalized model and reduce overfitting of the data. In some embodiments, the regularization includes an LI or L2 penalty. For example, in some embodiments, the one or more second models 136 is a plurality of logistic regression models, in which a first subset of the plurality of logistic regression models is trained using a regularization that includes an LI penalty on lower and upper parameters and a second subset of the plurality of logistic regression models is trained using a regularization that includes an L2 penalty on lower and upper parameters.

[0215] In some embodiments, model training of the one or more second models includes regularization that comprises spatial regularization (e.g., determined based on a priori and / or experimental knowledge) or dropout regularization. In some embodiments, the regularization comprises penalties that are independently optimized.

[0216] Generally, training the one or more second models comprises updating the plurality of parameters (e.g., weights) for each of the one or more second models in response to such training data. In some embodiments this is done through backpropagation (e.g., gradient descent). First, a forward propagation is performed, in which input data (the plurality of cellular constituent abundance values for each respective training compound in a plurality of training compounds) is accepted one by one into each of the one or more second models, and the model output for each of these training compounds from each of the one or more second models is calculated based on the initial set of model parameters (e.g., weights, parameters 132) for each of the one or more second models. In some embodiments, prior to training, the model parameters (e.g., weights and / or hyperparameters) for each of the one or more second models are randomly assigned (e.g, initialized). In some embodiments, model parameters are transferred from a previously saved plurality of parameters or from a pretrained model (e.g., by transfer learning).

[0217] In some embodiments, a backward pass is then performed by calculating an error gradient for each of the one or more second models, where the error is determined by calculating a loss (e.g., error) based on model outputs of each of the one or more second models (e.g., the predicted DILI risk) and the actual known DILI risk for each trainingcompound. Parameters 138 (e.g., weights) are then updated for each of the one or more second models 136 by adjusting parameter values based on the calculated loss, thereby training the one or more second models.

[0218] For example, in some general embodiments of machine learning, backpropagation is a method of training a plurality of weights of a model. The output of the untrained model (e.g., the predicted DILI risk) is first generated using a set of arbitrarily selected initial weights upon inputting the plurality of cellular constituent abundance values. The output (predicted DILI risk) is then compared with the original input (e.g., the actual DILI risk) for each training compound by evaluating an error function to compute an error (e.g., using a loss function). The model weights are then updated such that the error is minimized (e.g., according to the loss function). In some embodiments, any one of a variety of backpropagation algorithms and / or methods are used to update the plurality of weights of each of the second models, as will be apparent to one skilled in the art.

[0219] In some embodiments, the loss function is mean square error, quadratic loss, mean absolute error, mean bias error, hinge, multi-class support vector machine, and / or crossentropy. In some embodiments, training the one or more second models comprises computing an error in accordance with a gradient descent algorithm and / or a minimization function. In some embodiments, training the one or more second models comprises computing a plurality of errors using a plurality of loss functions. In some embodiments, each loss function in a plurality of loss functions receives a same or a different weighting factor.

[0220] In some embodiments, the error function is used to update one or more parameters (e.g., weights) in each of the one or more second models by adjusting the values of one or more parameters in each of the one or more second models by amounts proportional to the calculated loss, thereby training the one or more second models. In some embodiments, the amount by which the parameters are adjusted is metered by a learning rate hyperparameter that dictates the degree or severity to which parameters are updated (e.g., smaller or larger adjustments). Thus, in some embodiments, the training updates all or a subset of the plurality of parameters of each of the one or more second models based on the learning rate. In some embodiments, the learning rate is a differential learning rate.

[0221] In some embodiments, the training process, including adjusting the plurality of parameters associated with each of the one or more second models (e.g., responsive to the difference between the predicted and actual DILI risks), is repeated for each training instance in a plurality of training instances.

[0222] In some embodiments, the plurality of training instances comprises at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 50, at least 100, at least 500, at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, or at least 7500 training instances. In some embodiments, the plurality of training instances comprises no more than 10,000, no more than 5000, no more than 1000, no more than 500, no more than 100, or no more than 50 training instances. In some embodiments, the plurality of training instances comprises from 3 to 10, from 5 to 100, from 100 to 5000, or from 1000 to 10,000 training instances. In some embodiments, the plurality of training instances falls within another range starting no lower than 3 training instances and ending no higher than 10,000 training instances.

[0223] In some such embodiments, the training includes repeating the adjustment of the parameters of each of the one or more second models (e.g., via backpropagation) over a plurality of training instances, therefore increasing each of the one or more second model’s accuracy in determining a compounds DILI risk.

[0224] In some embodiments, the model training trains the one or more second models through a first evaluation of an error function. In some such embodiments, the model training trains the one or more second models following a first updating of one or more parameters based on a first evaluation of an error function. In some alternative embodiments, the model training trains after each of at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 100, at least 500, at least 1000, at least 10,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, or at least 1 million evaluations of an error function. In some such embodiments, the model training trains the one or more second models through at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 100, at least 500, at least 1000, at least 10,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, or at least 1 million updatings of one or more parameters based on the at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 100, at least 500, at least 1000, at least 10,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, or at least 1 million evaluations of an error function.

[0225] In some embodiments, the model training is repeated until the one or more second models individually, collectively, or collectively in conjunction with the one or more first models satisfies a minimum performance requirement. For example, in some embodiments,the model training is repeated until the error calculated for the one or more second models (either individually, collectively, or in conjunction with the one or more first models), following evaluation of the error functions (e.g., differences between predicted and actual DILI risk) satisfies an error threshold. In some embodiments, the error calculated by the error function satisfies an error threshold when the error is less than 20 percent, less than 18 percent, less than 15 percent, less than 10 percent, less than 5 percent, or less than 3 percent.

[0226] In an example embodiment, the one or more second models is trained using a categorical cross-entropy loss in a multi-task formulation, in which each covariate in a plurality of covariates corresponds to a cost function in plurality of cost functions and each respective cost function in the plurality of cost functions has a common weighting factor.

[0227] Referring to block 262, in some embodiments, the plurality of second models is between two and 1000 logistic regression models. Logistic regression models are disclosed in Agresti, An Introduction to Categorical Data Analysis, 1996, Chapter 5, pp. 103-144, John Wiley & Son, New York, which is hereby incorporated by reference.

[0228] Referring to block 264, in some embodiments, the one or more second models comprises 10 or more second modules, 20 or more second modules, 30 or more second modules, 40 or more second modules, 50 or more second modules, 60 or more second modules, 70 or more second modules, 80 or more second modules, or 90 or more second modules.

[0229] Block 266. Referring to block 266, in some embodiments, each second model in the one or more second models is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model. Details of such models are disclosed above in the definitions section.

[0230] In some embodiments, each second model comprises a plurality (e.g., 10 or more, 20 or more, 30 or more, 50 or more, 100 or more, 200 or more, 300 or more, 500 or more, 1000 or more, or 10,000 or more) of parameters.

[0231] In some embodiments, each second model comprises a plurality of parameters (e.g., weights and / or hyperparameters). In some embodiments, the plurality of parameters for each second model comprises at least 10, at least 50, at least 100, at least 500, at least 1000, at least 2000, at least 5000, at least 10,000, at least 20,000, at least 50,000, at least 100,000, at least 200,000, at least 500,000, at least 1 million, at least 2 million, at least 3 million, at least 4 million or at least 5 million parameters. In some embodiments, the plurality of parametersfor each second model comprises no more than 8 million, no more than 5 million, no more than 4 million, no more than 1 million, no more than 500,000, no more than 100,000, no more than 50,000, no more than 10,000, no more than 5000, no more than 1000, or no more than 500 parameters. In some embodiments, the plurality of parameters for each second model comprises from 10 to 5000, from 500 to 10,000, from 10,000 to 500,000, from 20,000 to 1 million, or from 1 million to 5 million parameters. In some embodiments, the plurality of parameters for each second model falls within another range starting no lower than 10 parameters and ending no higher than 8 million parameters.

[0232] In some embodiments, the training of each second model is further characterized by one or more hyperparameters (e.g., one or more values that may be tuned during training). In some embodiments, the hyperparameter values are tuned (e.g., adjusted) during training. In some embodiments, the hyperparameter values are determined based on the specific elements of a training dataset and / or one or more inputs (e.g., cells, cellular constituent modules, covariates, etc.). In some embodiments, the hyperparameter values are determined using experimental optimization. In some embodiments, the hyperparameter values are determined using a hyperparameter sweep. In some embodiments, the hyperparameter values are assigned based on prior template or default values.

[0233] In some embodiments, a respective hyperparameter of the one or more hyperparameters comprises a learning rate. In some embodiments, the learning rate is at least 0.0001, at least 0.0005, at least 0.001, at least 0.005, at least 0.01, at least 0.05, at least 0.1, at least 0.2, at least 0.3, at least 0.4, at least 0.5, at least 0.6, at least 0.7, at least 0.8, at least 0.9, or at least 1. In some embodiments, the learning rate is no more than 1, no more than 0.9, no more than 0.8, no more than 0.7, no more than 0.6, no more than 0.5, no more than 0.4, no more than 0.3, no more than 0.2, no more than 0.1 no more than 0.05, no more than 0.01, or less. In some embodiments, the learning rate is from 0.0001 to 0.01, from 0.001 to 0.5, from 0.001 to 0.01, from 0.005 to 0.8, or from 0.005 to 1. In some embodiments, the learning rate falls within another range starting no lower than 0.0001 and ending no higher than 1. In some embodiments, the one or more hyperparameters further include regularization strength (e.g., L2 weight penalty, dropout rate, etc.). For instance, in some embodiments, each second model (e.g, a neural network) is trained using a regularization on a corresponding parameter (e.g, weight) of each hidden neuron in the plurality of hidden neurons. In some embodiments, the regularization includes an LI or L2 penalty.

[0234] In some embodiments, a respective hyperparameter of the one or more hyperparameters for a second model is the identity of a loss function. In some embodiments,the loss function is mean square error, flattened mean square error, quadratic loss, mean absolute error, mean bias error, hinge, multi-class support vector machine, and / or crossentropy. In some embodiments, the loss function is a gradient descent algorithm and / or a minimization function.

[0235] In some embodiments, each second model is associated with one or more activation functions (e.g., in instances where such models are neural networks). In some embodiments, an activation function in the one or more activation functions is tanh, sigmoid, softmax, Gaussian, Boltzmann-weighted averaging, absolute value, linear, rectified linear unit (ReLU), bounded rectified linear, soft rectified linear, parameterized rectified linear, average, max, min, sign, square, square root, multiquadric, inverse quadratic, inverse multiquadric, polyharmonic spline, swish, mish, Gaussian error linear unit (GeLU), and / or thin plate spline. In some embodiments, each second model outputs a predicted DILI risk responsive to the inputting of the cellular constituent abundance values for a particular concentration 116 with a particular exposure time 118 into the model.

[0236] Block 268. Referring to block 268, in some embodiments, the first value and the one or more second values are used to determine a risk that the test chemical compound will induce the liver injury in the human subject. In such embodiments, the one or more first model and the one or more second models act jointly as an ensemble model in which the one or more first value from the one or more first models and the one or more second values form the one or more second models determine a risk that the test chemical compound will induce the liver injury in the human subject.

[0237] In some embodiments, this ensemble model comprises at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 20, at least 30, at least 40, at least 50, at least 60, at least 70, at least 80, at least 90, at least 100, at least 200, at least 300, at least 400, or at least 500 component models (e.g., one or more first models and the remainder independently trained instances of the second model).

[0238] In some embodiments, the ensemble model comprises no more than 500, no more than 400, no more than 300, no more than 200, or no more than 100 component models (e.g., one or more first models and the remainder independently trained instances of the second model). In some embodiments, the ensemble model comprises no more than 100, no more than 50, no more than 40, no more than 30, or no more than 20 component models. In some embodiments, the ensemble model comprises between 1 and 50, between 2 and 20, between 5 and 50, between 10 and 80, between 5 and 15, between 3 and 30, between 10 and 500, between 2 and 100, or between 50 and 100 component models. In some embodiments, theensemble model comprises another range of component models starting no lower than 2 component models and ending no higher than 500 component models.

[0239] In some embodiments, the output of the ensemble model is formed by combining a plurality of outputs obtained from the plurality of component models (e.g., one or more first values from the one or more first models and one or more second values from the one or more second models). In some embodiments, the plurality of outputs from the one or more first models and the one or more second models is combined using any measure of central tendency known in the art, including but not limited to a mean, median, mode, weighted mean, weighted median, weighted mode, arithmetic mean, midrange, midhinge, trimean, and / or Winsorized mean. For instance, the final determination from the ensemble model can be obtained based on the average of the outputs across all component models in the ensemble model. In some embodiments, the output of each of the one or more first models and the one or more second models is a DILI probability having a value of between 0 and 1.0. In some such embodiments, the DILI probability outputted by each model of the one or more first models and each model of the one or more second models is combined using any measure of central tendency known in the art.

[0240] In some embodiments, the plurality of outputs is combined using a voting method. For example, in some embodiments, the plurality of outputs is combined by tallying the number of outputs from each component model in the ensemble model, that indicate that the test compound is a DILI risk at a particular concentration. In some embodiments, the plurality of outputs from the component models is combined using a majority vote at a particular concentration. In some embodiments, each component model in the ensemble model is unweighted (e.g., each component model has one vote in the ensemble model). In some embodiments, one or more component models in the ensemble model is further weighted (e.g., has greater than 1 vote in the ensemble model).

[0241] In some embodiments the method described above in conjunction with Figure 2 is used to evaluate a plurality of test compounds for DILI. In such embodiments, each test compound in the plurality of test compounds is run through the method of Figure 3. Thus, if there are 100 test compounds, in such embodiments, method 300 is run 100 times, where each instance in the 100 times is for a different one of the test compounds.

[0242] In some embodiments, the plurality of test compounds includes at least 5, at least 10, at least 15, at least 20, at least 30, at least 40, at least 50, at least 100, at least 200, at least 300, at least 400, at least 500, at least 800, at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, at least 8000, at least 10,000, at least 20,000, at least 30,000, at least50,000, at least 80,000, at least 100,000, at least 200,000, at least 500,000, at least 800,000, at least 1 million, or at least 2 million test compounds

[0243] In some embodiments, the plurality of compounds includes no more than 10 million, no more than 5 million, no more than 1 million, no more than 500,000, no more than 100,000, no more than 50,000, no more than 10,000, no more than 8000, no more than 5000, no more than 2000, no more than 1000, no more than 800, no more than 500, no more than 200, or no more than 100 test compounds. In some embodiments, the plurality of compounds consists of from 10 to 500, from 100 to 10,000, from 5000 to 200,000, or from 10,000 to 1 million test compounds.

[0244] In some embodiments, the plurality of test compounds is between 10 and 1 x 106test compounds. In some embodiments, the plurality of test compounds is between 100 and 100,000 test compounds. In some embodiments, the plurality of test compounds is between 1000 and 100,000 test compounds.

[0245] Referring to block 270, in some embodiments, the risk is in the form of a probability that the test chemical compound will induce the liver injury in the human subject. In some embodiments the risk is a value between 0 and 1 where zero indicates that there is little to no chance that the test chemical compound will induce the liver injury in the human subject and one indicates a 100 percent chance that the test chemical compound will induce the liver injury in the human subject.

[0246] Blocks 272-276. Referring to block 272, in some embodiments, the risk is one level of risk from among an enumerated set of levels of risk that the test chemical compound will induce the liver injury in the human subject. Referring to block 274, in some such embodiments, the enumerated set of levels of risk consists of a low, a medium, and a high level of risk that the test chemical compound will induce the liver injury in the human subject.

[0247] Referring to block 276, in some embodiments, the risk is in the form of the lowest concentration of the test chemical compound predicted by the one or more first models and the one or more second models to induce the liver injury in the human subject.

[0248] Block 278. Referring to block 278, in some embodiments, the method further comprises determining a safety margin of the test chemical compound using the lowest concentration of the test chemical compound predicted by the one or more first models and the one or more second models to induce the liver injury in the human subject (e.g., determining the safety margin comprises dividing the risk by a Cmax or an LDso of the test chemical compound). In some embodiments in accordance with block 278, the first valueand the one or more second values of the ensemble model are used to determine a risk that the test chemical compound will induce the liver injury in the human subject at each of several different concentrations thereby producing a predicted DILI probability dose response curve such as the one shown in Figure 6. This predicted dose response curve is then used to compute the safety margin for the test chemical compound. For instance, the predicted dose response curve is used to estimate the LD50 of the test chemical compound, i.e., the dose at which the predicted DILI probability exceeds 50%. Furthermore, the separate pharmacokinetic study is used to determine the Cmax concentration of the test chemical compound. The DILI safety margin is calculated as the fold-change between LDso and Cmax. In some embodiments, when the estimated safety margin is greater than 50 the test chemical compound is considered to not be a DILI risk chemical. In some embodiments, when the calculated safety margin is less than 50 the test chemical compound is considered to be a DILI risk chemical. In some embodiments, a threshold other than 50 is used to determine whether the compound has a suitable safety margin. For example, in some embodiments a threshold of 10 or 25 is used. In other embodiments a threshold of 30, 35, 40, 45, 55, 60, 65, or 70 is used instead.

[0249] Block 280. Referring to block 280, in some embodiments, the method informs the selection of one or more human subjects for treatment with the test chemical compound and / or selection of one or more human subjects for continuation or discontinuation of treatment with the test chemical compound. For instance, in some embodiments, cells from each subject in a population of subject are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118. Then cellular constituent abundance values 122 and pathway module values from such exposed cells are inputted into the ensemble model (comprising the one or more first models 130 and the one or more second models 136) in order to determine a safety margin for the test chemical compound in such subjects using the techniques disclosed above with respect to block 278 and only those subjects that have a safety margin of greater than 50 are treated with the test chemical compound. In some embodiments, a threshold other than 50 is used to determine whether a particular subject has suitable safety margin. For example, in some embodiments a threshold of 30, 35, 40, 45, 55, 60, 65, or 70 is used instead to determine whether or not a particular subject can use the test chemical compound to treat a disease afflicting the subject.

[0250] Referring to block 282, in some embodiments, the method informs the dosing amount, duration, and / or frequency of the test chemical compound in one or more human subjects for treatment. For instance, in some embodiments, cells from a single subject orfrom a population of subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118. Then cellular constituent abundance values 122 and pathway module values from such exposed cells are inputted into the ensemble model (comprising the one or more first models 130 and the one or more second models 136) in order to determine a safety margin for the test chemical compound in the single subject or the population of subjects using the techniques disclosed above with respect to block 278 and only those subjects that have a safety margin of greater than 50 are treated with the test chemical compound. In some embodiments, a threshold other than 50 is used to determine whether a subject has suitable safety margin. For example, in some embodiments a threshold of 30, 35, 40, 45, 55, 60, 65, or 70 is used instead to determine whether or not a subject can use the test chemical compound to treat a disease afflicting the human subject.

[0251] Blocks 284-288. Referring to block 284, in some embodiments, the method informs the design of a clinical trial, the clinical trial comprising the use of the test chemical compound. Referring to block 286, in some such embodiments, the method informs the type and / or level of acceptable adverse events, one or more inclusion criteria, and / or one or more exclusion criteria for the clinical trial. Referring to block 288, in some such embodiments, the method informs one or more modifications of the clinical trial while the clinical trial is in progress. In one example cells from candidate subjects for the clinical trial subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118. Then the cellular constituent abundance values 122 and pathway module values from such exposed cells are inputted into the ensemble model (comprising the one or more first models 130 and the one or more second models 136) in order to determine a suitable dosage of the test chemical compound in the trial. In another example, cells from candidate subjects for the clinical trial subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118 in order to determine whether such candidate subject can tolerate the test chemical compound and therefore can participate in the trial. In still another example, cells from candidate subjects for the clinical trial subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118 in order to determine whether such candidate subject can remain in the trial.

[0252] Blocks 290-294. Referring to block 290, in some embodiments, the method informs the design of an adaptive clinical trial, the adaptive clinical trial comprising the use of the test chemical compound. Referring to block 292, in some embodiments, the method informs the type and / or level of acceptable adverse events, one or more inclusion criteria, and / or one or more exclusion criteria for the adaptive clinical trial. Referring to block 294, insome embodiments, the method informs one or more modifications of the adaptive clinical trial while the clinical trial is in progress. In one example cells from candidate subjects for the adaptive clinical trial subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118. Then the cellular constituent abundance values 122 and pathway module values from such exposed cells are inputted into the ensemble model (comprising the one or more first models 130 and the one or more second models 136) in order to determine a suitable dosage of the test chemical compound in the adaptive trial. In another example, cells from candidate subjects for the clinical trial subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118 in order to determine whether such candidate subject can tolerate the test chemical compound and therefore can participate in the adaptive trial. In still another example, cells from candidate subjects for the clinical trial subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118 in order to determine whether such candidate subject can remain in the adaptive trial.

[0253] Referring to block 296, in some embodiments, the method informs classification of the test chemical compound as a compound suitable for human therapy and / or testing or a compound unsuitable human therapy and / or testing. For instance, in some embodiments, cells from a population of subjects are exposed to the test chemical compound at one or more concentration 116 / exposure time pairs 118. Then cellular constituent abundance values 122 and pathway module values from such exposed cells are inputted into the ensemble model (comprising the one or more first models 130 and the one or more second models 136) in order to determine a safety margin for the test chemical compound the population of subjects using the techniques disclosed above with respect to block 278. When the compound is determined to have a safety margin of greater than 50 across the tested population of subjects (e.g., the measure of central tendency of the safety margin across the population of subjects) the test chemical compound is determined to be suitable for human therapy and / or testing. When the compound is determined to have a safety margin of less than 50 across the tested population of subjects (e.g., the measure of central tendency of the safety margin across the population of subjects) the test chemical compound is determined to not be suitable for human therapy and / or testing. In some embodiments, a threshold other than 50 is used to determine whether a subject has suitable safety margin. For example, in some embodiments a threshold of 30, 35, 40, 45, 55, 60, 65, or 70 is used instead to determine whether or not the test chemical compound is suitable for human therapy and / or testing.

[0254] Referring to block 298, in some embodiments, the method further comprises formulating the test chemical compound for use in a therapy. In some embodiments the therapy is to alleviate a condition such as inflammation. In some embodiments the therapy is to alleviate or treat a disease or disorder. In some embodiments the disease or disorder is cancer, a hematologic disorder, an autoimmune disease, an inflammatory disease, an immunological disorder, a metabolic disorder, a neurological disorder, a genetic disorder, a psychiatric disorder, a gastroenterological disorder, a renal disorder, a cardiovascular disorder, a dermatological disorder, a respiratory disorder, a viral infection, or other disease or disorder.

[0255] In some embodiments formulating the test chemical compound for use in a therapy comprises manufacturing a composition comprising the test chemical compound and one or more excipient and / or one or more pharmaceutically acceptable carrier and / or one or more diluent.

[0256] Such excipients and / or carriers include all conventional solvents, dispersion media, fillers, solid carriers, coatings, antifungal and antibacterial agents, dermal penetration agents, surfactants, isotonic and absorption agents and the like. It will be understood that the compositions of the present disclosure may also include other supplementary physiologically active agents.

[0257] An exemplary carrier is pharmaceutically “acceptable” in the sense of being compatible with the other ingredients of the composition (e.g., the composition comprising the test chemical compound) and not injurious to a subject. The compositions may conveniently be presented in unit dosage form and may be prepared by any methods well known in the art of pharmacy. Such methods include the step of bringing into association the active ingredient with the carrier that constitutes one or more accessory ingredients. In general, the compositions are prepared by uniformly and intimately bringing into association the active ingredient with liquid carriers or finely divided solid carriers or both, and then if necessary shaping the product.

[0258] Exemplary compounds, compositions or combinations of the present disclosure (e.g., the test chemical compound) formulated for intravenous, intramuscular or intraperitoneal administration, or a pharmaceutically acceptable salt, solvate or prodrug thereof may be administered by injection or infusion.

[0259] Injectables for such use can be prepared in conventional forms, either as a liquid solution or suspension or in a solid form suitable for preparation as a solution or suspension in a liquid prior to injection, or as an emulsion. Carriers can include, for example, water,saline (e.g., normal saline (NS), phosphate-buffered saline (PBS), balanced saline solution (BSS)), sodium lactate Ringer's solution, dextrose, glycerol, ethanol, and the like; and if desired, minor amounts of auxiliary substances, such as wetting or emulsifying agents, buffers, and the like can be added. Proper fluidity can be maintained, for example, by using a coating such as lecithin, by maintaining the required particle size in the case of dispersion and by using surfactants.

[0260] The compound, composition or combinations of the present disclosure (e.g., the test chemical compound) may also be suitable for oral administration and may be presented as discrete units such as capsules, sachets or tablets each containing a predetermined amount of the active ingredient; as a powder or granules; as a solution or a suspension in an aqueous or non-aqueous liquid; or as an oil-in-water liquid emulsion or a water-in-oil liquid emulsion. The active ingredient may also be presented as a bolus, electuary or paste.

[0261] A tablet may be made by compression or molding, optionally with one or more accessory ingredients. Compressed tablets may be prepared by compressing in a suitable machine the active ingredient (e.g., the test chemical compound) in a free-flowing form such as a powder or granules, optionally mixed with a binder (e.g., inert diluent, preservative disintegrant (e.g. sodium starch glycolate, cross-linked polyvinyl pyrrolidone, cross-linked sodium carboxymethyl cellulose) surface-active or dispersing agent). Molded tablets may be made by molding in a suitable machine a mixture of the powdered compound moistened with an inert liquid diluent. The tablets may optionally be coated or scored and may be formulated so as to provide slow or controlled release of the active ingredient therein using, for example, hydroxypropylmethyl cellulose in varying proportions to provide the desired release profile. Tablets may optionally be provided with an enteric coating, to provide release in parts of the gut other than the stomach.

[0262] The compound, composition or combinations of the present disclosure (e.g., the test chemical compound) may be suitable for topical administration in the mouth including lozenges comprising the active ingredient in a flavored base, usually sucrose and acacia or tragacanth gum; pastilles comprising the active ingredient in an inert basis such as gelatine and glycerin, or sucrose and acacia gum; and mouthwashes comprising the active ingredient in a suitable liquid carrier.

[0263] The compound, composition or combinations of the present disclosure (e.g., the test chemical compound) may be suitable for topical administration to the skin may comprise the compounds dissolved or suspended in any suitable carrier or base and may be in the form of lotions, gel, creams, pastes, ointments and the like. Suitable carriers include mineral oil,propylene glycol, polyoxyethylene, polyoxypropylene, emulsifying wax, sorbitan monostearate, polysorbate 60, cetyl esters wax, cetearyl alcohol, 2-octyldodecanol, benzyl alcohol and water. Transdermal patches may also be used to administer the compounds of the invention.

[0264] The compound, composition or combination of the present disclosure (e.g., the test chemical compound) may be suitable for parenteral administration include aqueous and non-aqueous isotonic sterile injection solutions which may contain anti-oxidants, buffers, bactericides and solutes which render the compound, composition or combination isotonic with the blood of the intended recipient; and aqueous and non-aqueous sterile suspensions which may include suspending agents and thickening agents. The compound, composition or combination may be presented in unit-dose or multi-dose sealed containers, for example, ampoules and vials, and may be stored in a freeze-dried (lyophilized) condition requiring only the addition of the sterile liquid carrier, for example water for injections, immediately prior to use. Extemporaneous injection solutions and suspensions may be prepared from sterile powders, granules and tablets of the kind previously described.

[0265] It should be understood that in addition to the active ingredients particularly mentioned above, the composition or combination of this present disclosure (e.g., the test chemical compound) may include other agents conventional in the art having regard to the type of composition or combination in question, for example, those suitable for oral administration may include such further agents as binders, sweeteners, thickeners, flavoring agents disintegrating agents, coating agents, preservatives, lubricants and / or time delay agents. Suitable sweeteners include sucrose, lactose, glucose, aspartame or saccharine. Suitable disintegrating agents include cornstarch, methylcellulose, polyvinylpyrrolidone, xanthan gum, bentonite, alginic acid or agar. Suitable flavoring agents include peppermint oil, oil of wintergreen, cherry, orange or raspberry flavoring. Suitable coating agents include polymers or copolymers of acrylic acid and / or methacrylic acid and / or their esters, waxes, fatty alcohols, zein, shellac or gluten. Suitable preservatives include sodium benzoate, vitamin E, alpha-tocopherol, ascorbic acid, methyl paraben, propyl paraben or sodium bisulphite. Suitable lubricants include magnesium stearate, stearic acid, sodium oleate, sodium chloride or talc. Suitable time delay agents include glyceryl monostearate or glyceryl di stearate.

[0266] III. ADDITIONAL EMBODIMENTS

[0267] Another aspect of the present disclosure provides a method of assessing a risk that a test chemical compound will induce a liver injury in a human subject. The methodcomprises obtaining, in electronic form, a cellular constituent dataset 112 comprising a plurality of cellular abundance values 122 for a plurality of cellular constituents 120 in a first plurality of cells that have been exposed to the test chemical compound for a first period of time. Details of such cellular constituent datasets are described above in conjunction with block 212.

[0268] The cellular constituent dataset is used to determine a respective activation value for each cellular constituent module in a plurality of cellular constituent modules, thereby obtaining a plurality of activation values. Each respective cellular constituent module in the plurality of cellular constituent modules comprises an independent subset of the plurality of cellular constituents. Each respective activation value of each respective cellular constituent module is a measure of central tendency of an abundance of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents, across the cellular constituent dataset. Each independent subset of the plurality of cellular constituents for each respective cellular constituent module comprises two or more cellular constituents. Each cellular constituent in the plurality of cellular constituents is in at least one independent subset of the plurality of cellular constituents associated with a cellular constituent module in the plurality of cellular constituent modules. In some embodiments the plurality of cellular constituents comprises 10 or more cellular constituents, and the plurality of cellular constituent abundance values comprises 1000 or more cellular constituent abundance values.

[0269] The plurality of activation values is inputted into a model thereby retrieving, as output from the model, a value indicative of a risk that the test chemical compound will induce the liver injury in the human subject.

[0270] In some embodiments, the first plurality of cells is primary human hepatocytes.

[0271] In some embodiments, the plurality of cellular abundance values is determined by single-cell ribonucleic acid (RNA) sequencing (scRNA-seq) data of the first plurality of cells. In some embodiments, the plurality of cellular abundance values is determined by bulk RNA sequencing. In some embodiments, the plurality of cellular abundance values is determined by scTag-seq. In some embodiments, the plurality of cellular abundance values is determined by single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq), CyTOF / SCoP, E-MS / Abseq, miRNA-seq, CITE-seq, or any combination thereof. Details of suitable cellular constituent abundance measurement techniques are disclosed above in conjunction with blocks 218 through 224.

[0272] In some embodiments the model is a logistic regression model. In alternative embodiments, the model is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model.

[0273] In some embodiments, the plurality of cellular constituent modules comprises 10 or more cellular constituent modules, 20 or more cellular constituent modules, 30 or more cellular constituent modules, 40 or more cellular constituent modules, 50 or more cellular constituent modules, 60 or more cellular constituent modules, 70 or more cellular constituent modules, 80 or more cellular constituent modules, or 90 or more cellular constituent modules.

[0274] In some embodiments, each cellular constituent in the plurality of cellular constituents is a particular gene, a particular mRNA associated with a gene, a carbohydrate, a lipid, an epigenetic feature, a metabolite, a protein, or a combination thereof. Further description of suitable cellular constituents is described above in conjunction with block 214.

[0275] In some embodiments, the method further comprises identifying the plurality of cellular constituent modules by a process in which one or more first datasets is obtained in electronic form. The one or more first datasets comprise, for each respective primary human hepatocyte cell in a plurality of primary human hepatocyte cells, where the plurality of primary human hepatocyte cells comprises twenty or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states, for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell, thereby accessing or forming a plurality of vectors. Each respective vector in the plurality of vectors (i) corresponds to a respective cellular constituent in the plurality of constituents and (ii) comprises a corresponding plurality of elements, each respective element in the corresponding plurality of elements representing an abundance of the respective cellular constituent in a corresponding primary human hepatocyte cell in the plurality of primary human hepatocyte cells. The plurality of vectors is clustered thereby forming a plurality of clusters, where each cluster in the plurality of clusters consists of an independent subset of the plurality of cellular constituents of a respective cellular constituent module in the plurality of cellular constituent modules.

[0276] In some alternative embodiments, the method further comprises identifying the plurality of cellular constituent modules by a process comprising obtaining one or more first datasets in electronic form. The one or more first datasets comprises, for each respectiveprimary human hepatocyte cell in a plurality of primary human hepatocyte cells, where the plurality of primary human hepatocyte cells comprises twenty or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states, for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell. A consensus cellular constituent module matrix is generated using the one or more first datasets using consensus non-negative matrix factorization. In this embodiment, the independent subset of the plurality of cellular constituents is identified for each respective cellular constituent module in the plurality of cellular constituent modules from the consensus matrix using multiple least squares regression of the corresponding abundances of the cellular constituents in the one or more first datasets against the consensus GEP usage matrix.

[0277] In some embodiments, an annotated cell state in the plurality of annotated cell states is an exposure of a cell in the second plurality of cells to a training compound under an exposure condition. In some embodiments, the exposure condition is a duration of exposure, a concentration of a training compound, or a combination of a duration of exposure and a concentration of a training compound.

[0278] In some embodiments, the plurality of annotated cell states includes exposure to a first training compound at a plurality of different concentrations. In some embodiments, the plurality of different concentrations range from a Cmax of the first training compound to an ICio of the first training compound. In some such embodiments, the plurality of annotated cell states includes exposure to a first training compound at a plurality of different durations of exposure.

[0279] In some embodiments, the plurality of annotated cell states includes exposure to each respective training compound in a plurality of training compounds at a corresponding plurality of different concentrations. In some such embodiments, the corresponding plurality of different concentrations range from a Cmax of each respective training compound to an ICio of the respective training compound.

[0280] In some embodiments, the plurality of annotated cell states includes exposure to each training compound in the plurality of training compounds at a plurality of different durations of exposure.

[0281] In some embodiments, the clustering is a form of graph clustering. In some such embodiments the graph clustering is Leiden clustering on a Pearson-correlation-based distance metric. In alternative embodiments, the graph clustering is Louvain clustering.

[0282] In some embodiments, the plurality of cellular constituent modules consists of between 10 and 2000 cellular constituent modules.

[0283] In some embodiments, the plurality of cellular constituents consists of between 100 and 35,000 cellular constituents.

[0284] Ranking of DILI predictive features.

[0285] As discussed herein, fold changes of single cell transcriptomic abundance data for a plurality of cellular constituents obtained from (i) cells that have been exposed to control (e.g., DMSO) versus (ii) cells that have been exposed to compound are obtained. These fold changes are used to determine pathway scores for a plurality of modules. Both the fold changes of the cellular constituents and the pathway scores serve as input features to an ensemble model. In particular, the plurality of pathway values is inputted into one or more first models of the ensemble model to retrieve one or more first model values. The fold changes of single cell transcriptomic abundance data is inputted into each model in one or more second models of the ensemble model to obtain as output from the one or more second models, one or more second values. The one or more first values and the one or more second values are used by the ensemble model to determine a risk that a test chemical compound will induce liver injury in the human subject. As such, in some embodiments of the present disclosure use the following input features: cellular constituent signatures (cellular constitute log-fold changes in abundance (DMSO vs. compound perturbation) and pathway module signatures: pathway module scores (e.g., -logl0(FDR)).

[0286] In some embodiments, each model of the one or more first models and each model of the one or more second models is a logistic regression. In some embodiments the one or more first models and the one or more second models collectively consists of between 70 models and 150 models. In some embodiments the one or more first models and the one or more second models collectively consists of between 90 models and 120 models.

[0287] In some embodiments, it is of interest to rank the importance of these input features. In some such embodiments, this is done by training the ensemble of models through Bagging (bootstrap aggregating), an ensemble learning technique that trains multiple models independently and combines them with equal weight. This helps reduce variance and prevent overfitting, leading to potentially better generalization. In some embodiments, the ensemble consists of a) multiple base models independently on different subsets of the training data, b) multiple base models with different hyperparameters, c) on gene-level and pathway-level signatures, each. Once trained, the feature importance for each feature across the ensemble is based on their coefficient (weights). Features with larger absolute coefficients areconsidered more important in determining the output. The ensemble feature importance is calculated as the average of coefficients of the ensemble models.

[0288] Ins some embodiments permutation importance is considered. For the top K features, in addition to feature importance, the values of one feature at a time are shuffled and the change in model performance (AUC) is measured. The larger the drop in performance, the more important the feature is.

[0289] IV. EXAMPLES

[0290] Provided herein are examples of assessing a risk that a test chemical compound will induce a liver injury in a human subject.

[0291] The Japanese Toxicogenomics Project (TGP) includes gene expression data, toxicological information and pathological data for 131 compounds screened in vitro for human toxicity. This data is available on the Internet at For each compound, the training datacontains the gene expression values (FARMS preprocessed), drug induced liver injury (DILI) and label classes (”- 1”,”+1”) for different time points (2 hours exposure, 8 hour exposure, and 24 hour exposure) and dose-levels (low, middle, and high).

[0292] For model training, 120 compounds were selected from the human TGP data at the high concentration. The training data is illustrated in Figure 4. Each column in Figure 4 represents one of the selected training compounds. The training compounds are organized from most DILI toxic at the left to least DILI toxic (no measured toxicity) at the right.

[0293] The abundance data of genes that are differentially expressed in the TGP data was enriched for known gene networks. Thus, each row in the bottom portion of the table in Figure 4 represents a different gene network (pathway module).

[0294] Table 2 below provides example pathway modules and the independent subset of genes associated with each such pathway module. It is to be noted that some genes are found in multiple pathways. In Table 2, OS stands for oxidative stress, LP stands for lipid peroxidation, MD stands for mitochondria dysfunction, and AD stands for antioxidant defenesns.

[0295] For any given compound the measurements of the differential genes associated with a pathway module will have a statistical significance as measured against the distribution of the expression value of randomly selected genes from the TGP dataset. This statistical significance can be expressed as a p-value. For a particular compound, a p-value of less than p < 0.05 is marked with a single asterisk (*), a p-value of less than < 0.01 is markedwith a double asterisk (**), and a p-value of less than < 0.001 is marked with a triple asterisk (***). The statistical significance (p-value) of each pathway module will vary from compound to compound. Table 2 shows the p-value of each of the pathway enrichment of predictive differentially expressed genes (DEGs) for at least one of the 120 compounds evaluated during training based on a statistical hypergeometric test to assess the significance of the observed enrichment of the pathway with DEGsTable 2. Example pathways and the independent subset of genes associated with each pathway.

[0296] In this example, each respective pathway module provides a respective pathway value for each respective compound that is a weighted average of the microarray expression data for the independent subset of genes associated with the respective pathway module upon exposure of TGP cell-cultures to the respective compound. In alternative embodiments, the respective pathway value for a given compound is the p-value for the pathway module with respect to the given compound.

[0297] An ensemble model was trained on the pathway values derived from the TGP data as well as the abundance data of the genes themselves in the TGP data. The ensemble model consisted of a first model and a plurality of second models. The pathway values upon exposure to each of the 120 training compounds was inputted into the first model. In this example, the first model was a standalone logistic regression model (first model). The output DILI predictions of the first model upon inputting the pathway values was trained against the known toxicity for the training compounds.

[0298] Separately, the gene expression data for 7254 genes from cells upon exposure to each compound in the set of 120 training compounds was inputted into each model in the plurality of second models. In this example, each of the second models was a separatelogistic regression model. The output DILI predictions of each of the second models upon inputting the gene expression values of the 7254 genes for each training compound was trained against the known toxicity values for the training compounds. One subset of the plurality of logistic regression models was trained on an LI basis and another subset of the plurality of logistic regression models was trained on an L2 basis.

[0299] For any given training compound at any given training composition, the output of the ensemble model was a range of DILI prediction probabilities, which is termed a confidence interval in this example. Contributing to this confidence interval was the DILI prediction probability of the first model and the DILI prediction probability of each model in the plurality of second models. Thus, for any given training compound, at any given concentration, the output of the ensemble model was a confidence interval of DILI prediction probabilities. Figure 5 shows the performance of this ensemble model (the first model together with the plurality of second models). In computing the performance for Figure 5, the DILI prediction probability for any given composition at any given composition exposure concentration was taken as the center of the measured confidence interval. For both graphs in Figure 5, the Y-axis is sensitivity (the true positive rate, TPR, which is calculated as TP / (TP+FN, where TP stands for true positive and FN stands for false negative). In the left hand graph of Figure 5, the X-axis is specificity (the false positive rate, FPR, which is calculated as FP / (FP+TN, where FP stands for false positive and TN stands for true negative). The left hand graph in Figure 5 shows that the trained ensemble model has an area under the curve (AUC) of 0.9. A model with AUC = 1 would discriminate compounds perfectly as having DILI toxicity or not. A model with an AUC = 0.5 would mean that the model has no apparent ability to determine whether compounds have DILI toxicity or not.

[0300] The right hand graph in Figure 5 shows the cut-point for the same ROC analysis. In this right hand graph, the X-axis is prediction probability. The cut-point in the right hand graph in Figure 5 is the point where the sensitivity curve and the specificity curve meet. The cut-off point determines the point at which the ensemble model classifies a drug as a DILI risk or not a DILI risk. By adjusting the cut-off point, one can trade-off between the true positive rate and the false positive rate. A lower cut-off (on the X-axis) classifies more instances as a DILI risk, leading to higher sensitivity but potentially higher false positives, while a higher threshold classifies fewer instances as positive, resulting in lower sensitivity but potentially lower false positives. After the classifier assigns prediction probabilities to a compound, a decision needs to be made about which class to assign to each compound. The decision is typically made by comparing the prediction probability to a predeterminedthreshold (e.g., the ROC cut-off point). If the prediction probability is above the threshold, the compound is classified as the positive class (i.e., the compound is deemed to be a DILI risk); otherwise, it is classified as the negative class (i.e., the compound is deemed to not be a DILI risk).

[0301] Validation using Hepatotoxicity Assay Methods.

[0302] The ensemble model trained using the TGP data was then tested against internal test (validation) data for thirty compounds. This internal validation data was SMART-Seq DE data of primary human hepatocytes upon exposure for 24 hours, at four different concentrations of each of the 30 compounds, ranging from Cmax to ICio, in three replicates. Primary human hepatocytes were used because of their physiologically relevance and predictive cell type for DILI.

[0303] Cell selection. Cryopreserved primary human hepatocytes, purchased from LifeNet Health, were selected using criteria for viability and plateability to meet standards for 96-well compatible culture. A specific donor was selected for long-term plateable (plateability at 10-15 days), 96 well compatible and demonstrate Grade A quality. Primary human hepatocytes were cultured according to the vendor’s protocols and proprietary media.

[0304] Experimental set-up.

[0305] On day zero, cells were seeded on Collagen I coated 96 well plates at a density of 0.5 million cells per mL using the vendor’s thawing and plating media.

[0306] On day one, cells were overlayed with 0.5mg / mL Matrigel until the next day to allow them to grow in an “ECM sandwich culture” condition, to promote a three-dimensional matrix for the hepatocytes.

[0307] On day two, cells were maintained with a refresh of the culture media to allow for stabilization of the culture.

[0308] On day three, cells were treated with test compounds. For the toxicology runs, compounds were solubilized in DMSO at lOOOmM, 500mM or 250mM dose (depending on solubility) according to the protocols described below. Compounds were further diluted 200x to bring down to final working concentration in Media, in a 6 point curve (full log down).

[0309] Solubilization of Test Compounds in DMSO

[0310] Relatively high concentrations of compounds in DMSO were utilized for the hepatotoxicity assays. To obtain optimal solubilization of compounds at high concentrations, a fit for purpose compound dissolution protocol was created, by which compounds weresolubilized on the day of the experiment to minimize potential “crushing out of solution” issues that might otherwise arise from freeze thaw cycles.

[0311] Stock solutions were ordered in powder form and dissolved at 200mM in sterile DMSO. At the instance when compounds were visually insoluble at 200mM, compounds were further diluted to a stock concentration at lOOmM, and still further diluted to 50mM if the compound was still not visibly soluble at lOOmM. For those instances where the compound was insoluble at 50mM, that compound was not used. Stock compounds were further processed according to two protocols (Toxicity assay to determine ICio, and Toxicity assay to acquire RNA samples).

[0312] Toxicity assay to determine ICio.

[0313] The stock compound solution was vortexed and placed in the bead bath to ensure maintenance of solubility.

[0314] Stock compound solutions were further processed by automation (Hamilton machine) to titrate down 10 fold 5 times in order to test a 6 point toxicity curve of each compound, each run in triplicate.

[0315] Stock compound solution was titrated 10 fold, samples are further diluted in hepatocyte maintenance media 200 fold to reach the final concentration of compound starting from lOOOpM, 500 pM, or 250 pM respectively of starting concentration. This final dilution of the compounds ensured a “safe” 0.5% DMSO concentration that was previously shown not to substantially induce toxicity in the cells nor to cause significant transcriptional activity.

[0316] Diluted solutions of test compounds were added to the cells at lOOpL volume via plates in 96 well format after aspirating the maintenance media that was on the cells prior to the addition of compounds.

[0317] Toxicity Assay to Acquire RNA Samples.

[0318] An assay procedure similar to that of the Toxicity Assay to Determine ICio was followed during the sequencing assay to acquire RNA samples where the only substantive difference was that starting concentrations of test compounds were predetermined upon results of the Toxicity Assay to Determine ICio mentioned above. ICio concentrations determined via the above assay were used as starting concentrations for the sequencing run to acquire RNA samples.

[0319] On day four, cells were assessed for viability using the LDH assay.

[0320] For the sequencing run: concentrations were selected in each toxicology run by defining the maximum tolerated dose (MTD). MTD was defined as the concentration atwhich <10% viability was observed. Compounds were again solubilized at 200mM, (100 or 50mM stock if not soluble at 200mM) in DMSO. Compounds were then diluted 200x in media to reach a final maximum tolerated dose and diluted in a 4-point curve in media (half log down).

[0321] On day four, cells were assessed for viability using the LDH assay, and sacrificed using Lysis buffer, to produce lysates that were then processed to extract RNA, which was then used to run SMART-Seq DE, which is a commercially available Smart-seq2 -based kit that is compatible with multiplex cDNA library preparation and allows improved throughput (Takara Bio).

[0322] Gene expression readout measured by SMART-Seq DE from perturbations (cells exposed to each composition) and DMSO (cells exposed to DMSO only) were processed using DESeq2, (https: / / genomebiology.biomedcentral.com / artic1es / 10.1186 / sl 3059-614- 0550-8), Love et al., 2014, “Moderated estimation of fold change and dispersion for RNA- seq data with DESeq2,” Genome Biol 15, 550 on the Internet at doi.org / 10.1186 / sl 3059-014- 0550-8 for normalization and log-fold change estimation of perturbation profiles against DMSO. DESeq2 was first used to estimate size factors for each sample. The size factor represents the scaling factor that should be applied to each sample to account for differences in sequencing depth. It was computed as the median of the ratios of gene counts to their geometric means across all samples. DESeq2 then uses these size factors to normalize the raw counts by dividing each gene's count in a sample by the corresponding size factor for that sample. This normalization step adjusts the counts for sequencing depth, making the samples comparable. Log-Fold Changes were then determined by taking the difference in the log2- transformed normalized counts (log2-counts) between perturbation versus DMSO for each gene. DESeq2 employed a shrinkage procedure, called “regularized log transformation,” to moderate log-fold changes. This helps improve the accuracy of log-fold changes for genes with low counts and high dispersion. Differential gene expression: Finally, DESeq2 performed statistical tests (Wald test) to determine if the observed log-fold changes were significantly different from zero. Thus the plurality of gene abundance values for the plurality of genes that have been exposed to the test chemical compound for a first period of time were in the form of gene expression log-fold changes (DMSO vs. compound perturbation). These log-fold changes were used to compute the pathway values for the one or more first models and were also inputted directly into the one or more second models in order to determine the risk that a test chemical compound would induce liver injury. Fromthis analysis, of the 30 compounds, 22 were determined to not be DILI save compounds whereas 8 were determined to be DILI safe compounds.

[0323] Against this validation data, the ensemble model exhibited 86% sensitivity (19 / 22 DILI) and 100% specificity (8 / 8 non-DILI) in one validation study. In a repeat of this validation study using data for the same 30 compounds in different laboratory settings, the ensemble model exhibited 89% sensitivity (17 / 19 DILI) and 100% specificity (9 / 9 non-DILI).

[0324] Of the 30 compounds, only three were misclassified: Labelatol, Acetaminophen, and Isoniazid. This compares to QSAR-based industry standards of 62% sensitivity and 75% specificity. Table 3 provides a listing, for each of the 30 validation compounds, of compound name, CMAX, toxicity dose, safety margin, and DILI status that is known from the literature.Table 3.

[0325] Figure 6 illustrates the validation results for Amiodarone. From Table 3, it is seen that the toxic dose for Amiodarone is 8 pM. The Cmax, a standard measurement in pharmacokinetics (the maximum or peak serum concentration that a drug achieves in a specified compartment or test area of the body after the drug has been administered and before the administration of a second dose), for Amiodarone is 0.8 pM. Therefore, the safety margin for Amiodarone is 10 and is deemed to be a DILI risk. The recommended safety margin for DILI risk is greater than 50. See Leishman et al.. 2020 “Revisiting the hERG safety margin after 20 years of routine hERG screening,” J Pharmaocol Toxicol Methods 105, 106900, which is hereby incorporated by reference.

[0326] Shown in Figure 6 are the predictions made by the ensemble model for primary human hepatocytes upon exposure for 24 hours, at each of four different concentrations of Amiodarone. Each of the four different concentrations is denoted by a black dot. Because the ensemble model includes a number of models that each predict a DILI probability, each dot is centered at the confidence interval defined by the minimum and maximum probability identified by any of the models within the ensemble model at the specified concentration. These confidence intervals are denoted by a vertical line through all DILI prediction probabilities through each of the models in the ensemble model. The average DILI prediction probability at each tested concentration was used to define curve 602. Using curve 602, the LDso is derived as the concentration at which the DILI probability exceeds 0.5. Thus Fig. 6 shows that, for Amiodarone, the LD50, given by a DILI probability of 0.5, is 7.9uM, while the Cmax is 1 ,34uM. Thus, the ensemble model predicts a DILI safety margin of 7.9 / 1.34 = 5.9, which is a value less than 50, and therefore the ensemble model classifies Amiodarone as a DILI risk, consistent with the known data for this compound provided in Table 3.

[0327] Figure 7 illustrates the expression gene signatures measured for 12 of the 30 test compounds. For each of the 12 compounds, there are four columns of expression data, arranged from lowest to highest test compound exposure concentration. In Figure 7, gene expression is in the form of a heat map from low gene expression (dark blue) to high gene expression (dark red). The ensemble model was unable to provide a DILI toxicity prediction for Erythromycin or Labetalol. Figure 7 shows that the gene expression signatures for these compounds is featureless, suggesting that the ensemble model was unable to read a signal from these gene signatures in order to predict DILI toxicity for these compounds.

[0328] Figures 8A and 8B illustrate the pathway values for 12 of the 30 test compounds for 16 different pathway modules. For each of the 12 compounds, there are four columns, arranged from the lowest to highest test compound exposure concentration. In Figure 8, the respective pathway value, for each respective pathway module, for each respective compound, is in the form of a heat map from low pathway value (dark blue) to high pathway value (dark red). Figure 8 also highlights the two compounds for which the ensemble model was unable to provide a DILI toxicity prediction: Erythromycin or Labetalol. Figure 8 shows that the pathway values upon exposure to these compounds is also featureless, suggesting that the ensemble model was unable to read a signal from these pathway module values in order to predict DILI toxicity for these compounds.

[0329] Figure 9 provides further explanation for while the ensemble model is unable to provide a prediction for Erthromycen and Labelatol. Exposure of primary human hepatocytes to either of these compounds does not cause expression of the known DILI markers listed in Figure 9.

[0330] Figure 10 plots the predicted safety margin for toxicity determined by the ensemble model (Y-axis) against the true DILI toxicity (X-axis) for the 30 test compounds. The ensemble model correctly labeled all but three of compounds that should have a safety margin below 50 because they are known to be DILI toxic. A safety margin below 50 is considered be at risk for DILI toxicity. The ensemble model only mis-labeled three of the compounds with a safety margin above 50 (DILI-safe). The ensemble model should have predicted a safety margin below 50 (DILI-toxic) for these three compounds. The three mislabeled compounds were Labelatol, Acetaminophen, and Isonazid.

[0331] Figure 11 illustrates how pre-clinical DILI safety margins predicted by the ensemble model correlate with known clinical therapeutic index. In Figure 11, the predicted safety margin for toxicity determined by the ensemble model (X-axis) is plotted against the known clinical therapeutic index (Y-axis) for the 30 test compounds. The correlation coefficient between (i) the predicted safety margin for toxicity determined by the ensemble model (X-axis) and (ii) the known clinical therapeutic index (Y-axis) for the 30 test compounds was 0.78. From Figure 11, it is seen that the compounds given the wrong DILI label by the ensemble model, Labelatol, Acetaminophen, and Isonazid, have a relatively large therapeutic index that is more similar to the therapeutic index of compounds that are DILI- safe.

[0332] This Example shows that the trained ensemble model can guide compound selection and optimization by de-risking DILI (drug-induced liver injury) through predictingDILI safety margins from in vitro transcriptomics with primary hepatocytes. This example shows that the trained ensemble model exhibited more predictive power than industry standard QSAR models. Moreover, the nature of the data used by the ensemble model, in vitro transcriptomics with primary hepatocytes upon compound exposure, is scalable.

[0333] REFERENCES CITED AND ALTERNATIVE EMBODIMENTS

[0334] All references cited herein are incorporated herein by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety for all purposes.

[0335] The present invention can be implemented as a computer program product that includes a computer program mechanism embedded in a non-transitory computer readable storage medium. For instance, the computer program product could contain the program modules shown in any combination of Figures 1-2. These program modules can be stored on a CD-ROM, DVD, magnetic disk storage product, or any other non-transitory computer readable data or program storage product.

[0336] The present description includes example systems, methods, techniques, instruction sequences, and computing machine program products that embody illustrative implementations. For purposes of explanation, numerous specific details are set forth in order to provide an understanding of various implementations of the inventive subject matter. It will be evident, however, to those skilled in the art that implementations of the inventive subject matter may be practiced without these specific details. In general, well-known instruction instances, protocols, structures and techniques have not been shown in detail.

[0337] Many modifications and variations of this invention can be made without departing from its spirit and scope, as will be apparent to those skilled in the art. The specific embodiments described herein are offered by way of example only. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. The invention is to be limited only by the terms of the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

WHAT IS CLAIMED IS:

1. A method of assessing a risk that a test chemical compound will induce a liver injury in a human subject, the method comprising:(A) obtaining, in electronic form, a cellular constituent dataset comprising a plurality of cellular constituent abundance values for a plurality of cellular constituents in a first plurality of cells that have been exposed to the test chemical compound for a first period of time;(B) using the cellular constituent dataset to determine a respective pathway value for each pathway module in a plurality of pathway modules, thereby obtaining a plurality of pathway values, wherein each respective pathway module in the plurality of pathway modules comprises an independent subset of the plurality of cellular constituents, each independent subset of the plurality of cellular constituents for each respective pathway module comprises two or more cellular constituents, each cellular constituent in the plurality of cellular constituents is in at least one independent subset of the plurality of cellular constituents associated with a pathway module in the plurality of pathway modules, the plurality of cellular constituents comprises 10 or more cellular constituents, and the plurality of differential cellular constituent abundance values comprises 500 or more cellular constituent abundance values;(C) inputting the plurality of pathway values into one or more first models thereby retrieving, as output from the first model, one or more first values;(D) inputting the plurality of cellular constituent abundance values into one or more second models thereby retrieving, as output from the one or more second models, one or more second values; and(E) the one or more first values and the one or more second values are used to determine a risk that the test chemical compound will induce the liver injury in the human subject.

2. The method of claim 1, wherein the first plurality of cells is primary human hepatocytes.

3. The method of claim 1 or 2, wherein the abundance of each cellular constituent in the plurality of cellular constituents in each cell in the first plurality of cells is determined by single-cell ribonucleic acid (RNA) sequencing (scRNA-seq) data of the first plurality of cells.

4. The method of claim 1 or 2, wherein the abundance of each cellular constituent in the plurality of cellular constituents in each cell in the first plurality of cells is determined by bulk RNA sequencing.

5. The method of claim 1 or 2, wherein the abundance of each cellular constituent in the plurality of cellular constituents in each cell in the first plurality of cells is determined by scTag-seq.

6. The method of claim 1 or 2, wherein the abundance of each cellular constituent in the plurality of cellular constituents in each cell in the first plurality of cells is determined by single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq), CyTOF / SCoP, E-MS / Abseq, miRNA-seq, CITE-seq, or any combination thereof.

7. The method of any one of claims 1-6, wherein each model in the one or more first models is a logistic regression model.

8. The method of claim 7, wherein each respective pathway value of each respective pathway module is a measure of central tendency of a cellular constituent abundance value of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents of the respective pathway.

9. The method of claim 7, wherein a respective pathway value of a respective pathway module in the plurality of pathway modules is a negative base 10 logarithm of a false discovery rate determined based on a distribution of cellular constituent abundances of each cellular constituent in the independent subset of cellular constituents corresponding to the respective subset of cellular constituents of the respective pathway module, relative to cellular constituent abundances of cellular constituents in the cellular constituent dataset.

10. The method of any one of claims 1-9, wherein the one or more second models is a single logistic regression model.

11. The method of any one of claims 1-9, wherein the one or more second models is a plurality of logistic regression models.

12. The method of claim 11, wherein a first subset of logistic regression models in the plurality of second models is a plurality of LI regression models, and a second subset of logistic regression models in the plurality of second models is a plurality of L2 regression models.

13. The method of any one of claims 1-12, wherein the plurality of second models is between two and 1000 logistic regression models.

14. The method of claim 1, wherein the one or more second models comprises 10 or more second modules, 20 or more second modules, 30 or more second modules, 40 or more second modules, 50 or more second modules, 60 or more second modules, 70 or more second modules, 80 or more second modules, or 90 or more second modules.

15. The method of claim 1, wherein each first model in the one or more first models is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model.

16. The method of claim 1, wherein each second model in the one or more second models is a regression module, a neural network model, a support vector machine model, a Naive Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a multinomial logistic regression model, a linear model, or a linear regression model.

17. The method of any one of claims 1-16, wherein each cellular constituent in the plurality of cellular constituents is a particular gene, a particular mRNA associated with a gene, a carbohydrate, a lipid, an epigenetic feature, a metabolite, a protein, or a combination thereof.

18. The method of any one of claims 1-17, the method further comprising identifying the plurality of pathway modules by a process comprising: obtaining one or more first datasets in electronic form, the one or more first datasets comprising: for each respective primary human hepatocyte cell in a plurality of primary human hepatocyte cells, wherein the plurality of primary human hepatocyte cells comprises twenty or more primary human hepatocyte cells and collectively represents a plurality of annotated cell states: for each respective cellular constituent in the plurality of cellular constituents: a corresponding abundance of the respective cellular constituent in the respective primary human hepatocyte cell, identifying the plurality of cellular constituents, in the one or more first datasets, as a plurality of differentially expressed cellular constituents across the one or more first datasets; and associating each independent subset of cellular constituents in the plurality of cellular constituents with a corresponding pathway for a pathway module in the plurality of pathway modules.

19. The method of 18, wherein an annotated cell state in the plurality of annotated cell states is an exposure of a cell in the plurality of primary human hepatocyte cells to a training compound under an exposure condition.

20. The method of claim 19, wherein the exposure condition is a duration of exposure, a concentration of a training compound, or a combination of a duration of exposure and a concentration of the training compound.

21. The method of any one of claims 18-20, wherein the plurality of annotated cell states includes exposure to a first training compound at a plurality of different concentrations.

22. The method of claim 21, wherein the plurality of different concentrations range from (i) a Cmax or an imputed Cmax of the first training compound to (ii) an ICio of the first training compound.

23. The method of any one of claims 18-20, wherein the plurality of annotated cell states includes exposure to a first training compound at a plurality of different durations of exposure.

24. The method of any one of claims 18-20, wherein the plurality of annotated cell states includes exposure to each respective training compound in a plurality of training compounds at a corresponding plurality of different concentrations.

25. The method of claim 24, wherein the corresponding plurality of different concentrations range from (i) a Cmax of the respective training compound to (ii) an ICio of the respective training compound.

26. The method of any one of claims 18-19, wherein the plurality of annotated cell states includes exposure to each training compound in the plurality of training compounds at a plurality of different durations of exposure.

27. The method of any one of claims 1-26, wherein the plurality of pathway modules consists of between 10 and 2000 pathway modules.

28. The method of any one of claims 1-27, wherein the plurality of cellular constituents consists of between 100 and 35,000 cellular constituents.

29. The method of any one of claims 1-28, wherein each independent subset of the plurality of cellular constituents for each respective pathway module comprises between two cellular constituents and three hundred cellular constituents.

30. The method of any one of claims 1-29, wherein the test chemical compound is an organic compound having a molecular weight of less than 2000 Daltons.

31. The method of any one of claims 1-30, wherein the test chemical compound is an organic compound that satisfies each of the Lipinski rule of five criteria.

32. The method of any one of claims 1-31, wherein the test chemical compound is an organic compound that satisfies at least three criteria of the Lipinski rule of five criteria.

33. The method of any one of claims 1-32, wherein the risk is in the form of a probability that the test chemical compound will induce the liver injury in the human subject.

34. The method of any one of claims 1-32, wherein the risk is one level of risk from among an enumerated set of levels of risk that the test chemical compound will induce the liver injury in the human subject.

35. The method of claim 34, wherein the enumerated set of levels of risk consists of a low, a medium, and a high level of risk that the test chemical compound will induce the liver injury in the human subject.

36. The method of any one of claims 1-32, wherein the risk in the form of the lowest concentration of the test chemical compound predicted by the first model and the one or more second models to induce the liver injury in the human subject.

37. The method of claim 36, the method further comprising: determining a safety margin of the test chemical compound using the lowest concentration of the test chemical compound predicted by the first model and the one or more second models to induce the liver injury in the human subject.

38. The method of claim 37, wherein the determining the safety margin comprises dividing the risk by a Cmax of the test chemical compound.

39. The method of any one of claims 1-38, wherein the liver injury is hepatitis, hepatotoxicityjaundice, hepatic fibrosis, cirrhosis, hepatomegaly, alcoholic and nonalcoholic liver fatty liver disease, and / or choluria.

40. The method of any one of claims 1-38, wherein the liver injury is characterized by an elevation of one or more liver enzymes, cholestasis, and / or alcoholic stools.

41. The method of any one of claims 1-40, wherein the method informs the selection of one or more human subjects for treatment with the test chemical compound and / or selection ofone or more human subjects for continuation or discontinuation of treatment with the test chemical compound.

42. The method of any one of claims 1-41, wherein the method informs the dosing amount, duration, and / or frequency of the test chemical compound in one or more human subjects for treatment.

43. The method of any one of claims 1-42, wherein the method informs the design of a clinical trial, the clinical trial comprising the use of the test chemical compound.

44. The method of any one of claims 1-42, wherein the method informs the design of an adaptive clinical trial, the adaptive clinical trial comprising the use of the test chemical compound.

45. The method of claim 43, wherein the method informs the type and / or level of acceptable adverse events, one or more inclusion criteria, and / or one or more exclusion criteria for the clinical trial.

46. The method of claim 44, wherein the method informs the type and / or level of acceptable adverse events, one or more inclusion criteria, and / or one or more exclusion criteria for the adaptive clinical trial.

47. The method of claim 43, wherein the method informs one or more modifications of the clinical trial while the clinical trial is in progress.

48. The method of claim 44, wherein the method informs one or more modifications of the adaptive clinical trial while the clinical trial is in progress.

49. The method of any one of claims 1-40, wherein the method informs classification of the test chemical compound as a compound suitable for human therapy and / or testing or a compound unsuitable human therapy and / or testing.

50. The method of any one of claims 1-49, wherein the method further comprises formulating the test chemical compound for use in a therapy.

51. The method of any one of claims 1-50, wherein each cellular constituent abundance value in the cellular constituent dataset represents a differential cellular constituent abundance between cell exposure to the test compound and cell exposure to a control.

52. The method of claim 51 wherein the control is dimethyl sulfoxide.

53. The method of any one of claims 1-51, wherein the one or more first models consists of a single first model.

54. The method of any one of claims 1-52, wherein each respective cellular constituent abundance value is in a form of a log-fold change of the corresponding cellular constituent relative to an abundance of the corresponding cellular constituent in a second plurality of cells exposed to a control solution for the first period of time.

55. The method of claim 54, wherein the control solution is dimethyl sulfoxide.

56. A computer system, comprising one or more processors and memory, the memory storing instructions for performing a method of assessing a risk that a test chemical compound will induce a liver injury or a cardiac injury in a human subject, the method comprising:(A) obtaining, in electronic form, a cellular constituent dataset comprising a plurality of cellular constituent abundance values for a plurality of cellular constituents in a first plurality of cells that have been exposed to the test chemical compound for a first period of time;(B) using the cellular constituent dataset to determine a respective pathway value for each pathway module in a plurality of pathway modules, thereby obtaining a plurality of pathway values, wherein each respective pathway module in the plurality of pathway modules comprises an independent subset of the plurality of cellular constituents, each independent subset of the plurality of cellular constituents for each respective pathway module comprises two or more cellular constituents,each cellular constituent in the plurality of cellular constituents is in at least one independent subset of the plurality of cellular constituents associated with a pathway module in the plurality of pathway modules, the plurality of cellular constituents comprises 10 or more cellular constituents, and the plurality of differential cellular constituent abundance values comprises 500 or more cellular constituent abundance values;(C) inputting the plurality of pathway values into one or more first models thereby retrieving, as output from the first model, one or more first values;(D) inputting the plurality of cellular constituent abundance values into one or more second models thereby retrieving, as output from the one or more second models, one or more second values; and(E) the one or more first values and the one or more second values are used to determine a risk that the test chemical compound will induce the liver injury in the human subject.

57. A computer system, comprising one or more processors and memory, the memory storing instructions for performing a method of assessing a risk that a test chemical compound will induce a liver injury or a cardiac injury in a human subject of any one of claims 1-55.

58. A non-transitory computer-readable medium storing one or more computer programs, executable by a computer, for assessing a risk that a test chemical compound will induce a liver injury or a cardiac injury in a human subject, the computer comprising one or more processors and a memory, the one or more computer programs collectively encoding computer executable instructions that perform a method comprising:(A) obtaining, in electronic form, a cellular constituent dataset comprising a plurality of cellular constituent abundance values for a plurality of cellular constituents in a first plurality of cells that have been exposed to the test chemical compound for a first period of time;(B) using the cellular constituent dataset to determine a respective pathway value for each pathway module in a plurality of pathway modules, thereby obtaining a plurality of pathway values, whereineach respective pathway module in the plurality of pathway modules comprises an independent subset of the plurality of cellular constituents, each independent subset of the plurality of cellular constituents for each respective pathway module comprises two or more cellular constituents, each cellular constituent in the plurality of cellular constituents is in at least one independent subset of the plurality of cellular constituents associated with a pathway module in the plurality of pathway modules, the plurality of cellular constituents comprises 10 or more cellular constituents, and the plurality of differential cellular constituent abundance values comprises 500 or more cellular constituent abundance values;(C) inputting the plurality of pathway values into one or more first models thereby retrieving, as output from the first model, one or more first values;(D) inputting the plurality of cellular constituent abundance values into one or more second models thereby retrieving, as output from the one or more second models, one or more second values; and(E) the one or more first values and the one or more second values are used to determine a risk that the test chemical compound will induce the liver injury in the human subject.

59. A non-transitory computer-readable medium storing one or more computer programs, executable by a computer, for assessing a risk that a test chemical compound will induce a liver injury or a cardiac injury in a human subject, the computer comprising one or more processors and a memory, the one or more computer programs collectively encoding computer executable instructions that perform a method of any one of claims 1-55.

Images