System and method for perturbational profiling of blood cell traits

Abstract

Genome wide association studies (GWAS) have demonstrated the power of genetics to detect previously unrecognized contributions to disease biology. To date one limitation has been the modest effect sizes of many of the detected loci, and the lack of a predictable relationship between the locus and specific mechanisms. These features have prevented the large-scale identification of new genetic targets for therapeutic intervention for all but a small number of GWAS loci. To directly address these limitations and build on the success of GWAS, the present disclosure presents a series of specific biological tools and analytics, which when combined with relevant clinical data at scale, enable the systematic discovery of previously unknown disease subsets, previously unknown disease genes of significant effect, previously unknown therapeutic targets, and disease-to-pharmacology matching at the individual patient level.

Description

SYSTEM AND METHOD FOR PERTURBATIONAL PROFILING OF BLOODCELL TRAITSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of and priority to U.S. Provisional Patent Application No. 63 / 543,274, filed on October 9, 2023, entitled “Perturbational Profiling,” the entire contents of which are herein incorporated by reference for all purposes.TECHNICAL FIELD

[0002] This disclosure relates to the field of blood sample and genetic data analysis, and more particularly to systems and methods for analyzing blood sample and genetic data from a subject to identify disease biomarkers, drug targets, and the like.BACKGROUND

[0003] Genome wide association studies (GWAS) have demonstrated the power of genetics to detect previously unrecognized contributions to disease biology. To date, limitations in comparative examples include the modest effect sizes of many of the detected loci and a lack of a predictable relationship between the locus and specific mechanisms. These features have prevented the large-scale identification of new genetic targets for therapeutic intervention for all but a small number of GWAS loci. Moreover, although GWAS have succeeded in discovering genetic risk loci linked to various disorders, the identification of underlying cellular biological mechanisms remains challenging due to the complex and heterogeneous nature of many common diseases.SUMMARY

[0004] To directly address these and other limitations and build on the success of GWAS, the present disclosure presents a series of specific biologic tools and analytics that, when combined with relevant clinical data at scale, enable the systematic discovery of previously unknown disease subsets, previously unknown disease genes of significant effect, previously unknown therapeutic targets, and disease-to-pharmacology matching at the individual patient level. The present disclosure establishes a framework, using peripheral bloodcells (e.g., of a human or another subject), chemical and pharmacological perturbations, and flow cytometry-based cell functional readouts to reveal latent cellular biological processes.

[0005] In examples as will be set forth herein, GWAS was performed based on these evoked cellular traits of up to 2600 subjects. Thereby, 119 genomic loci were identified implicating 96 genes that are associated with these blood cell responses. Deep clinical phenotypic information was integrated and associations between the evoked blood cellular phenotypes and subsets of chronic diseases (in examples, cardiometabolic syndromes) were identified. Using an evoked neutrophil trait as an example, the present disclosure describes a population of pro-inflammatory anti-apoptotic neutrophils that is present in subjects with specific subsets of cardiometabolic disease. Comparatively simple multigenic models based on this trait predicted the risk of developing chronic kidney disease in Type 2 diabetes patients, for example. By expanding the phenotypic space for GWAS, the systems and methods of the present disclosure are able to identify genetic variants linked to large-effect response differences, stratify subjects, and efficiently characterize underlying biology .

[0006] According to one aspect of the present disclosure, a method of analyzing a blood sample comprises evoking a latent blood cellular phenotype, including adding at least one of a chemical compound or a biological perturbation to the blood sample; assaying the evoked blood cellular phenotype with at least one assay; generating a disease association model based on the assayed blood cellular phenotype and on a disease outcome record, thereby to output a biomarker for an associated disease or disease subset; and using the at least one assay, performing a chemical screen on blood cells to identify a therapeutic compound that modulates the biomarker.

[0007] According to another aspect of the present disclosure, a method of analyzing a blood sample comprises evoking a latent blood cellular phenotype, including adding at least one of a chemical compound or a biological perturbation to the blood sample; assaying the evoked blood cellular phenotype with at least one assay; performing a genome wide association study (GWAS) on the evoked blood cellular phenotype, thereby to generate a druggable target for the associated disease or disease subset; using the at least one assay, performing a chemical screen on blood cells to identify a therapeutic compound that modulates the druggable target.BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Some embodiments of the disclosure are described herein with reference to the accompanying figures. The description, together with the figures, makes apparent to a personhaving ordinary skill in the art how some embodiments of the disclosure may be practiced. The figures are for the purpose of illustrative discussion and no attempt is made to show structural details of an example in more detail than is necessary for a fundamental understanding of the teachings of the disclosures. In the drawings:

[0009] FIG. 1A illustrates an example of the use of chemical perturbations to expand the phenotypic space of quantitative blood profiles in accordance with various aspects of the present disclosure.

[0010] FIG. IB illustrates an example of the use of chemical perturbations to expand the phenotypic space of quantitative blood profiles in accordance with various aspects of the present disclosure.

[0011] FIG. 1C illustrates an example of the use of chemical perturbations to expand the phenotypic space of quantitative blood profiles in accordance with various aspects of the present disclosure.

[0012] FIG. ID illustrates an example of the use of chemical perturbations to expand the phenotypic space of quantitative blood profiles in accordance with various aspects of the present disclosure.

[0013] FIG. 2A illustrates examples of blood cell distributions under baseline and perturbation conditions in accordance with various aspects of the present disclosure.

[0014] FIG. 2B illustrates examples of blood cell distributions under baseline and perturbation conditions in accordance with various aspects of the present disclosure.

[0015] FIG. 3 illustrates example projections of blood-response readouts in accordance with various aspects of the present disclosure.

[0016] FIG. 4A illustrates example distance correlations between blood trait readout pairs in accordance with various aspects of the present disclosure.

[0017] FIG. 4B illustrates example distance correlations between blood trait readout pairs in accordance with various aspects of the present disclosure.

[0018] FIG. 4C illustrates example distance correlations between blood trait readout pairs in accordance with various aspects of the present disclosure.

[0019] FIG. 4D illustrates example distance correlations between blood trait readout pairs in accordance with various aspects of the present disclosure.

[0020] FIG. 5A illustrates example non-overlapping regions with nearby candidate genes in accordance with various aspects of the present disclosure.

[0021] FIG. 5B illustrates example non-overlapping regions with nearby candidate genes in accordance with various aspects of the present disclosure.

[0022] FIG. 6A illustrates a subset of example pairwise association in accordance with various aspects of the present disclosure.

[0023] FIG. 6B illustrates a subset of example pairwise association in accordance with various aspects of the present disclosure.

[0024] FIG. 7A illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0025] FIG. 7B illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0026] FIG. 7C illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0027] FIG. 7D illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0028] FIG. 7E illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0029] FIG. 7F illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0030] FIG. 7G illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0031] FIG. 7H illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0032] FIG. 71 illustrates example pairwise association in accordance with various aspects of the present disclosure.

[0033] FIG. 8A illustrates examples of neutrophil death and relationship with neutrophil pro-mflammatory responses in accordance with various aspects of the present disclosure.

[0034] FIG. 8B illustrates examples of neutrophil death and relationship with neutrophil pro-mflammatory responses in accordance with various aspects of the present disclosure.

[0035] FIG. 9 illustrates an example link between neutrophil population emergence and heat exposure in accordance with various aspects of the present disclosure.

[0036] FIG. 10 illustrates an example neutrophil response in accordance with various aspects of the present disclosure.

[0037] FIG. 11A illustrates an example relationship between genes and regulation of neutrophils in accordance with various aspects of the present disclosure.

[0038] FIG. 11B illustrates an example relationship between genes and regulation of neutrophils in accordance with various aspects of the present disclosure.

[0039] FIG. 11C illustrates an example relationship between genes and regulation of neutrophils in accordance with various aspects of the present disclosure.

[0040] FIG. 12A illustrates an example set of polygenic scores in accordance with various aspects of the present disclosure.

[0041] FIG. 12B illustrates an example set of polygenic scores in accordance with various aspects of the present disclosure.

[0042] FIG. 12C illustrates an example set of polygenic scores in accordance with various aspects of the present disclosure.

[0043] FIG. 12D illustrates an example set of polygenic scores in accordance with various aspects of the present disclosure.

[0044] FIG. 12E illustrates an example set of polygenic scores in accordance with various aspects of the present disclosure.

[0045] FIG. 12F illustrates an example set of polygenic scores in accordance with various aspects of the present disclosure.

[0046] FIG. 13A illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0047] FIG. 13B illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0048] FIG. 13C illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0049] FIG. 13D illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0050] FIG. 13E illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0051] FIG. 13F illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0052] FIG. 13G illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0053] FIG. 13H illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0054] FIG. 131 illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0055] FIG. 13J illustrates an example set of hazard ratio estimates in accordance with various aspects of the present disclosure.

[0056] FIG. 14 illustrates an example response in accordance with various aspects of the present disclosure.

[0057] FIG. 15 illustrates an example assessment of activities in accordance with various aspects of the present disclosure.

[0058] FIG. 16A illustrates an example decision tree analysis in accordance with various aspects of the present disclosure.

[0059] FIG. 16B illustrates an example decision tree analysis in accordance with various aspects of the present disclosure.

[0060] FIG. 17 illustrates an example method of analyzing blood in accordance with various aspects of the present disclosure.

[0061] FIG. 18 illustrates an example method of analyzing blood in accordance with various aspects of the present disclosure.

[0062] FIG. 19 illustrates an example blood analysis system in accordance with various aspects of the present disclosure.

[0063] FIG. 20 illustrates example association curves in accordance with the present disclosure.DETAILED DESCRIPTION

[0064] In the following detailed description, reference is made to the accompanying drawings in which specific examples are shown by way of illustration. These examples are described in sufficient detail to enable those of ordinary skill in the art to practice the disclosure.It should be understood, however, that the detailed description and the specific examples, while indicating examples of embodiments of the disclosure, are given by way of illustration only and not by way of limitation. From this disclosure, various substitutions, modifications, additions rearrangements, or combinations thereof within the scope of the disclosure may be made and will become apparent to those of ordinary skill in the art.

[0065] Unless otherwise indicated, the various features illustrated in the drawings may not be drawn to scale. The illustrations presented herein are not necessarily intended to be actual views of any particular method, device, or system, but are merely idealized representations that are employed to describe various embodiments of the disclosure. Accordingly, the dimensions of the various features as illustrated may be arbitrarily expanded or reduced for clarity. In addition, some of the drawings may be simplified for clarity. Thus, the drawings may not depict all of the components of a given apparatus (e.g., device) or method. In addition, like reference numerals may be used to denote like features throughout the specification and figures.

[0066] It should be understood that any reference to an element herein using a designation such as “first,” “second,” and so forth does not limit the quantity or order of those elements, unless such limitation is explicitly stated. Rather, these designations may be used herein as a convenient method of distinguishing between two or more elements or instances of an element. Thus, a reference to first and second elements does not mean that only two elements may be employed there or that the first element must precede the second element in some manner. Also, unless stated otherwise a set of elements may comprise one or more elements.

[0067] Unless otherwise specified or indicated by context, the terms “a,” “an,” and “the” mean “one or more.” As used herein, unless otherwise limited or defined, “or” indicates a non-exclusive list of components or operations that can be present in any variety of combinations, rather than an exclusive list of components that can be present only as alternatives to each other. For example, a list of “A, B, or C” indicates options of: A; B; C; A and B; A and C; B and C; and A, B, and C. Correspondingly, the term “or” as used herein is intended to indicate exclusive alternatives only when preceded by terms of exclusivity, such as “only one of,” or “exactly one of.” For example, a list of “only one of A, B, or C” indicates options of: A, but not B and C; B, but not A and C; and C, but not A and B. In contrast, a list preceded by “one or more” (and variations thereon) and including “or” to separate listed elements indicates options of one or more of any or all of the listed elements. For example, the phrases “one or more of A, B, or C” and “at least one of A, B, or C” indicate options of: oneor more A; one or more B; one or more C; one or more A and one or more B; one or more B and one or more C; one or more A and one or more C; and one or more A, one or more B. and one or more C. Similarly, a list preceded by “a plurality of’ (and variations thereon) and including “or” to separate listed elements indicates options of one or more of each of multiple of the listed elements. For example, the phrases “a plurality of A, B, or C” and “two or more of A, B, or C” indicate options of: one or more A and one or more B; one or more B and one or more C; one or more A and one or more C; and one or more A, one or more B, and one or more C.

[0068] As used herein, “about,” “approximately,” “substantially,” and “significantly” will be understood by persons of ordinary skill in the art and will vary to some extent on the context in which they are used. If there are uses of these terms which are not clear to persons of ordinary skill in the art given the context in which they are used, “about” and “approximately” will mean plus or minus <10% of the particular term and “substantially” and “significantly” will mean plus or minus >10% of the particular term.

[0069] As used herein, the terms “include” and “including” have the same meaning as the terms “comprise” and “comprising” in that these latter terms are “open” transitional terms that do not limit claims only to the recited elements succeeding these transitional terms. The term “consisting of,” while encompassed by the term “comprising,” should be interpreted as a “closed” transitional term that limits claims only to the recited elements succeeding this transitional term. The term “consisting essentially of,” while encompassed by the term “comprising,” should be interpreted as a “partially closed” transitional term which permits additional elements succeeding this transitional term, but only if those additional elements do not materially affect the basic and novel characteristics of the claim.

[0070] As used herein, the term “subject” may be used interchangeably with the term “patient” or “individual” and may include an “animal” and in particular a “mammal.” Mammalian subjects may include humans and other primates, domestic animals, farm animals, and companion animals such as dogs, cats, guinea pigs, rabbits, rats, mice, horses, cattle, cows, and the like.

[0071] As used herein, the terms “treat” or “treatment” encompass both “preventative” and “curative” treatment. “Preventative” treatment is meant to indicate a postponement of development of a disease, a symptom of a disease, or medical condition, suppressing symptoms that may appear, or reducing the risk of developing or recurrence of a disease or symptom. “Curative” treatment includes reducing the severity of or suppressing the worsening of anexisting disease, symptom, or condition. Thus, treatment includes ameliorating or preventing the worsening of existing disease symptoms, preventing additional symptoms from occurring, ameliorating or preventing the underlying systemic causes of symptoms, inhibiting the disorder or disease, e.g., arresting the development of the disorder or disease, relieving the disorder or disease, causing regression of the disorder or disease, relieving a condition caused by the disease or disorder, or stopping the symptoms of the disease or disorder.

[0072] Precision medicine stnves to reclassify complex heterogenous diseases or states of wellness into distinct biologically defined groups, thereby enabling targeted therapies and improved disease outcomes. Examples include the subdivision of common cancers by somatic driver mutations, the discovery of eosinophilic variants of asthma, and the recognition that some presentations of heart failure may arise from the accumulation of amyloidogenic proteins, which in turn can be subdivided further into subtypes based on the aggregated protein. The realization of precision medicine has been hindered by the lack of readily available measures of discrete biological pathway activities in most common diseases. Historical approaches have focused on mining large patient biobanks combining archived DNA, RNA, and serum or plasma samples with associated clinical records. Although such comparative examples have identified common genetic variants associated with traditional clinical outcomes, they typically have not been successful at capturing the relevant underlying cell biology, limiting their utility in producing mechanistic insights and therapeutic implications.

[0073] The present disclosure establishes a framework that bridges genetic variants and complex diseases through standardized phenotyping of primary subject cells. In examples, circulating subject blood cells (e.g., circulating human blood cells) are utilized, as these reflect an integral of many physiological processes, disease states, and a host of environmental factors, including active therapies. For example, dysregulation of the primary functions of hematopoiesis can result in disease progression via mechanisms such as the contribution of inflammation to atherosclerosis and insulin resistance, or the hyperactivity in coagulation in pathological thrombosis. In addition to circulating cells with their repertoire of responses, blood plasma contains hormones, secreted proteins, and metabolites that can carry signals to blood cells or other cell types such as endothelial cells. There are numerous examples of peripheral blood phenotypes that represent the fundamental cell biology of discrete tissuespecific pathophysiologic defects, suggesting that peripheral blood may offer a generalizable diagnostic window into multiple organ systems and integrative physiol ogy.

[0074] Comparative GWAS studies on whole blood samples primarily focused on complete blood counts (CBC), clinical parameters describing these blood cell counts, volumes, and distribution. By studying these basic parameters in large cohorts under baseline conditions, the genetic architecture of hematopoiesis and blood diseases has been mapped in detail, other comparative work expanded the measured phenotypes to include flow cytometry-derived parameters with the aim of better describing cellular function. The human functional genome project profiled cytokine responses and baseline immune parameters in approximately 500 subjects. However, these comparative studies did not consider the dynamic responses of blood cells to environmental conditions, which likely contribute to their effects on disease development, progression, and prevention.

[0075] The present disclosure is based, at least in part, on a hypothesis that treating whole blood ex vivo with diverse stressors or stimuli would enable the identification of latent differential cellular responses and previously unknown disease-associated endophenotypes. It was anticipated that this expansion of latent phenotypic space would evoke traits determined by large effect size common alleles enabling efficient target identification and improving the prediction of incident events. Moreover, given that biological pathways are reused across diverse tissues and organ systems, insights from whole blood may be of relevance to a wider range of conditions originating in different tissues. By identifying intermediate cellular phenotypes, subcategories of disease and specific pathophysiologic mechanisms that can be targeted more directly were sought.

[0076] The methodologies set forth herein are based on several underlying components, including unbiased strategies for detecting evoked (and previously latent) phenotypes in individual subjects, using latent phenotypes to identify previously unknown subsets of disease, identifying and validating the genetic basis for latent phenotypes where this is present, refining comparative disease definitions and training previously unknown disease definitions using latent phenotypes as biomarkers, defining and validating druggable subject genetic targets for these disease definitions, and direct screening for initial therapeutic compounds in clinical samples.

[0077] By combining, in individual subjects, large-scale chemical screens (perturbations) and cellular or physiologic assays (of any type as the approach is agnostic to the assay or context and may be in vitro, ex vivo and in vivo) it is possible to evoke previously inaccessible or undetectable (latent) phenotypes, at large scale (hundreds or thousands of compounds and hundreds of phenotypes). For example, this approach is generalizable to anyassay using primary cells from peripheral blood, including morphometry, physiology, specific functions, protein, gene expression, and the like before and after a perturbation. The present disclosure includes a showing that these phenotypes are not only evident in human cells but can reproducibly be translated to animal models (see below) thus further accelerating human to bench and back to human movement of discovery and development work.

[0078] The particular perturbations within the platform can be directly controlled for focused discovery strategies including a) the identification of environmental or acquired factors modulating disease, b) the extension of this concept to nutrition, c) the identification of extreme responders in heterogeneous populations, and / or d) matching of drugs or nutrients to individual subjects. Another example which highlights the ability to study gene-environment interaction at the individual subject level is the use of microbiome intermediates as perturbations to define their acute or chronic effects and the biological basis of these effects. These assays also enable high resolution exploration of the biology at scale.

[0079] The approach generates large numbers of previously unknown testable intermediate phenotypes (or metrics for measuring health, disease, reserve or response); including parameters that can be measured to diagnose diseases or disease subsets, disease mechanisms or treatment endpoints. These assays (ranging from cell biology to whole organism physiolog)') allow the rigorous mechanistic characterization of previously inaccessible or hidden biology which can then be efficiently advanced to understand mechanism as a result of the large effect sizes compared with most human common allele effect sizes. As noted above translation at a genetic, phenotypic or chemical level is readily feasible to model organisms.

[0080] Existing diseases are defined in terms of combinations of prior data (clinical phenotypes) with particular predictable medical outcomes. However, the scale of clinical phenotyping has changed little in the last few decades. In fact, comparative genetic studies are dependent on clinical collections which were initiated 25-30 years ago. As the platform set forth herein enables a step-wise expansion in common intermediate phenotypes when combined with modem analytics in the electronic medical record, or other datasets, it is possible to train these phenotypes against outcomes of interest. In the first instance, simple models are used to reclassify existing diagnoses, but it is also feasible to train models against much more specific outcomes.

[0081] Exploiting the availability of the phenotypes discussed herein and the existing common disease annotation in a clinical record, it is possible to efficiently identify discretesubsets of disease and to then confirm that these phenotypes and diseases are genetically linked and explore mechanism in human cells or in animal models.

[0082] Any of these biological assays can also be used in rigorous genetic studies to identify relevant genetic causes if these exist. The approach reproducibly identifies previously latent and inaccessible genetic differences which can then be tested for association and causal involvement in clinical different outcomes or treatments in specific populations. The genetic results facilitate at least two steps.

[0083] The first step includes identification of the mechanism of the intermediate phenotype. Since the effect sizes for many of these intermediate phenotypes are large, the latency of the phenotypes can be defined in the general population, and there is a direct experimental system to test causality (the phenotypes themselves), it is possible to identify the underlying genes efficiently using common allele genotyping in only a few hundred individuals. The phenotypes themselves (both basal and evoked states) translate to model organisms. The second step includes independent validation of the causal contribution of previously unknown phenotypes to specific disease states. By defining the genetic basis for the intermediate trait in terms of a single or multiple interacting loci, it is then possible to independently validate whether the phenotype is truly an intermediate for the disease of interest by testing whether the genetic loci determining the phenoty pe are also predictive of the disease in an independent cohort (e g., MGB Biobank or UK Biobank).

[0084] These steps also enable definitive assessment of the effect size, and are thus useful for selecting potential therapeutic targets. Notably, even where there are no genetically mediated differences, these phenotypes have the potential to reclassify disease subtypes and treatment response phenotypes in ways that not only allow the discovery of additional targets but are directly aligned with improving responses to existing therapies, previously unknown actionable outcomes, and / or emerging therapies.

[0085] Comparative clinical phenotypic endpoints often do not have the relevant resolution for either care delivery' or genetic use. Adding previously unmeasured metrics has the potential to redefine disease and biology, while also making these elements more computable for care delivery and discovery or translation between the two. These efforts include a comprehensive analysis of the matrix of previously unknown phenotypes against discrete objective outcomes of acute medical interest. This requires dissociation of the subjective components in the EMR from the objective. The present disclosure presents multiple approaches to in silico phenotyping which integrate objective and subjective data streams tocreate computational models of disease-relevant outcomes such as time to specific preventable clinical events or development of specific therapeutic complications.

[0086] One potential purpose of the tools and their integration set forth herein is to identify human genetic targets for specific subsets of disease. This requires a showing that the phenotypes identified and the genes controlling these phenotypes are both independently associated with the disease subset of interest. This approach allows one to select from large numbers of genotype:: phenotype pairs for those with the largest signal and most important disease associations. Then, as a result of the accessibility of the cellular phenotype and its established genetics one can validate the cell biology in patients and the genetics in an independent set of subjects. It will be shown in detail below that this is highly reproducible and rapidly generates human genetic targets.

[0087] It is also possible to use those phenotypes that are identified and have proven are of disease relevance to undertake formal chemical screens in primary human cells from patients. Typically, drug discovery requires the generation of in vitro cellular assays and animal models to identify potential lead compounds and test these, but these efforts can take a decade and the issues of human relevance remain. The present disclosure shows that it is possible to move the current targets to chemical screens in the transitional ways using either target driven screens or phenotype driven screens, but to expand the utility of the approach the increased scale of the assays (several hundred compounds per patient sample per day) is exploited to enable the direct execution of cellular phenotype driven screens in human disease samples. This has resulted in the definition of scalable chemical screens in ex vivo primary human cellular assays expressly designed for this purpose. This aspect has the potential, in combination with the data above, to disrupt the strategies and timelines in comparative techniques drug discovery and development. For example, the present disclosure shows multiple compounds identified in a screen performed in primary human samples. It is also feasible to execute the entire process using highly specific chemical probes to define previously unknown uses for compounds interacting with particular targets and this is readily executed in humans with the scalable screening approaches developed herein.

[0088] There have not been comparative efforts to add new dynamic cellular phenotypes to human disease biology, so the approach set forth herein immediately adds thousands of unique disease-related phenotypes which are then trained using Al against specific clinical disease outcomes. It is also shown that these can be selected for their disease relevance, and previously unknown targets identified quickly because of effect size (achallenge with comparative GWAS where this has only been possible with 4-5 out of 500000 examples). The present disclosure also presents the chemical results from an actual screen in human primary cells as direct proof of concept. This takes Al driven drug discovery from cell lines to actual human samples. It should be noted that the present disclosure presents only a subset of examples, and thus the examples set forth below should be interpreted as illustrative and not exhaustive or limiting.

[0089] Methodology

[0090] FIGS. 1 A-D illustrate an example of the use of chemical perturbations to expand the phenotypic space of quantitative blood profdes, and illustrates that blood cell responses are associated with clinical phenotypes and genetic variants. In a clinical setting, whole blood cytometry may be used to quantify circulating cells as part of standardized diagnostic tests (e.g., CBC). In support of the methodologies set forth herein, a whole blood cytometry analyzer (Sysmex XN-1000) was adapted to systematically profde peripheral blood from over 4700 study participants under 36 conditions, more than 2600 subjects were genotyped, and genomewide association studies were performed for all blood-perturbation profiles. This is illustrated in image (a) of FIG. 1A, which shows a recruitment setting and application of a hematological analyzer for CBC together with perturbation agents to systematically measure cellular responses in whole blood samples across the clinical cohort.

[0091] Side scatter (SSC), forward scatter (FSC), and side fluorescence (SFL) of blood cells was recorded using four different fluorescent dyes including WDF, WNR, RET, and PLT- F, each of which quantifies morphological and intracellular properties of specific blood cell populations. Chemical stressors evoked distinct cellular states for blood cells that were not typically observed under baseline conditions and led to the detection of previously unknown cell populations in the three-dimensional cytometry measurements. FIGS. 2A-B illustrate this, and in particular shows blood cell distributions under baseline and perturbation conditions. In FIGS. 2A-B, the distribution of blood cytometry readouts is shown for baseline conditions and three perturbation conditions. Aggregate counts for each bin were calculated across all available samples for each condition and normalized to the number of cells measured in each channel.

[0092] Cellular gates were then defined based on the empiric distributions of blood cells under perturbation conditions, and defined parameter sets were derived for all observed cell populations. This is illustrated in image (b) of FIG. 1 A. In particular, image (b) shows data- driven gate definitions, corresponding to known and previously unknown cellular states inresponse to perturbation conditions. The initial perturbation conditions were selected to represent discrete classes of exposure likely to contribute to blood cell responses: i) to simulate physiological stressors (e.g., infection and sepsis using Pam3CSK4 and LPS; inflammasome activation using Nigericin), ii) chemical stressors (e.g., lithium chloride, alhydrogel), iii) gut microbiome metabolites (e.g. acetic acid, isobutyric acid), and iv) drugs with known mechanisms of action such as ciproflaxin and empagliflozin. These results are illustrated in Table 1.TABLE 1 - Perturbation Conditions

[0093] Up to 36 condition-specific blood responses were recorded for each patient, and quantitative profiles characterizing each cell population using cell counts were calculated, as well as median and standard deviations for SSC, FSC, and SFL parameters for each blood cell populations. These results are presented in FIG. IB. Cell gates were used to derive highdimensional quantitative readouts for 278 blood cell parameters across 37 environmental conditions including inflammatory stimuli (LPS, Pam3CSK4), heat or approved and experimental compounds (Dapagliflozin, Empagliflozin, Captopril). Each perturbation condition was measured for up to 3300 subjects. Compared to the baseline condition, each perturbational condition evoked particular changes in the characteristics of different blood lineages, resulting in a series of distinct cellular profiles. FIG. 3 illustrates a projection of blood-response readouts. In FIG. 3, quantitative parameters were defined by first gating cytometry readouts by cell type and calculating statistical parameters such counts, or median and standard deviation in dimensions such as SSC and SFL for each cell type. With these chemical perturbations, quantification for each subject was expanded from 278 parameters to more than 4000 parameters on average, greatly expanding the phenotypic space that could be interrogated.

[0094] Across the 36 perturbation conditions, 650-3300 subjects were measured per condition. Blood response profiles were them systematically associated with clinical traits including quantitative lab values and diagnostic codes to identify the clinical endpoints and disease syndromes that are reflected in the evoked blood response readouts. Blood parameters and response to perturbation conditions were associated with clinical phenotypes such as ICD10 diagnostic codes and lab measurements, as illustrated in FIG. 1C. Genetic loci associated with blood-perturbation responses were also identified, which were often specific to perturbation conditions, cell populations and physical readouts. The perturbation screening setting yielded many genetic associations which were specific to blood cell types andenvironmental stimuli. By comparing similar conditions in the same cohort, fine-grained comparisons between perturbation conditions and specific associated blood parameters were possible, as illustrated in FIG. ID.

[0095] When comparing blood-response profiles, the perturbation conditions, readouts and associated genetic loci formed clusters of related conditions and cell types, suggesting the evoked blood profiles are informative for specific biological processes. This is further illustrated in FIGS. 4A-D, which collectively show distance correlations between all pairs of blood trait readouts with significant genetic associations projected into a 2-dimensional embedding using UMAP. Each trait is assigned a color by the cell type, the type of readout, associated candidate genes at the GWAS locus and the perturbation condition.

[0096] Studying perturbation-blood cell responses among subjects in the study cohort, genetic variants associated with these responses were identified. Linear, univariate associations of 278 cellular phenotypes in 36 different conditions were tested against >3.5 million imputed variants in 260-2200 study subjects that were genotyped. Variants with high LD were clumped to identify more than 100 genomic loci that were significantly associated with at least two blood-perturbation readouts, as illustrated in Table 2.Table 2 - Demographics of Study Cohort

[0097] Next, 48 unique, non-overlapping regions with nearby candidate genes were identified. FIGS. 5A-B illustrate these regions, which were used to compare to comparative genetic studies of blood traits. Genome-wide significant associations with p-value < 5e-8 colored by perturbation condition (FIG. 5A) and cell type (FIG. 5B). Large circles indicate associated variants with nearby genes (coding variants, introns, regulatory regions), and small circles indicate variants located in intergenic regions. For clarity, only a subset of loci is shown for readouts with many significant associations. Table 3 presents an ovendew of traits, cell types, candidate genes and previously reported blood trait associations.Table 3 - Genetic Regions Associated with Whole-Blood Perturbation Response Traits

[0098] In Table 3, associations for blood traits and perturbation conditions were clumped to produce unique genomic regions across multiple conditions. SNPs with lowest p- value for each clumped region were selected as lead SNPs. The trait names contain the channel, condition, and readout, for example WDF Empa 1 5h NE3 CV SFL indicates a readout in the WDF channel, with empagliflozin treatment, quantifying the SFL coefficient of variation (CV) of a neutrophil subpopulation (NE3).

[0099] Approximately half of the identified associations (25 out of 48 genetic regions with candidate genes) had been described as blood biomarker associations under baseline conditions with parameters that are part of a CBC study encompassing 170,000 to over 500,000subjects. Previously unknown associations in previously unreported cell types were observed for many of the previously reported loci (12 out of 25), such as WBC responses associated with SLC83A3 whereas previously only reticulocyte-based associations had been described. Additionally, 23 previously unknown genetic regions associated with blood cell response to perturbations were identified that have not been previously described, for example, the response to high dose empagliflozin associated with variants in TMC04. This gene had been associated with chronic inflammatory diseases. Most associations observed were specific to a particular blood lineage, such as numerous reticulocyte readouts associated with TRIM58 or neutrophil-specific associations with PFKP and ACSL1.

[0100] The addition of chemical stressors to evoke latent blood cell readouts and increase response differences among subjects (see FIGS. 2A-B and 3) also made it possible to identify robust genetic associations with small sample sizes. Treatment with compounds of known biological effect yielded reproducible blood trait associations that were absent at baseline, even with much larger cohort sizes. For example, neutrophil and multiple other WBC responses induced by inflammatory stimuli such as Pam3CSK4 or LPS showed a strong association with a missense variant at the TLR1 locus (e.g., rs5743618, wdf Pam3CSK4 19h NE1 Med FSC, p-value=8.2E-103, n=1300). This association between TLR1 and WBC traits was not described previously in cohorts studying CBC parameters with over half a million subjects. The same SNP in TLR1 has been linked with asthma and allergic diseases through unclear mechanisms. These results suggest a potential role for neutrophil biology in mediating these disease phenotypes.

[0101] To test whether perturbation-based blood cell traits reflect individuals’ disease status, systematic tests were performed for associations between blood readouts and a collection of structured phenotypes based on EHR data, including diagnostic codes and lab measurements. The diagnostic status for multiple common disorders was significantly associated with variation in blood-perturbation readouts. These associations are illustrated in FIGS. 6A-B and 7A-I.

[0102] FIGS. 6A-B show pairwise association between blood readouts and clinical lab values or diagnostic codes. Association t-scores were estimated using linear and logistic regression models for quantitative lab measurements and binary traits, respectively. Positive associations are shown in red, and negative associations are shown in blue. Associations for only a subset of blood traits are shown in FIGS. 6A-B, whereas FIGS. 7A-I show results for all blood traits with significant genetic associations. FDR was calculated for all pairwiseassociation models (327 blood readouts with genetic hits and 43 clinical traits or lab values). Significant associations with FDR < 0.01 are indicated by points. Table 4 shows ICD10 diagnostic codes used for clinical trait definitions.Table 4 - ICD10 Diagnostic Code Definitions Used for Disease Associations

[0103] Notably, the addition of perturbations elicited unique disease associations that were absent at baseline. For example, neutrophil variability in SFL at baseline (wdf Baseline NE4 SD SFL) showed no significant association with disease. However, the same parameter under 21h Alhydrogel perturbation (wdf Alhydrogel 21h NE4 SD SFL) showed negative associations with multiple cardiometabolic diseases, including heart failure (cases=494, t=-3.1, padj=8.6E-3), T2D (cases=847, t=-7.0, padj=1.7E-10), as well as chronic kidney disease (cases=453, t=-3.0, padj=1.2E-2). Certain blood readouts were also observed to show an association with very specific disease phenotypes, such as platelet variability in SFL under Colchicine 20h (ret Colchicine 20h PLT CV SFL, n=199, t=-3.9, padj=7.6E-4), KC1 17h (ret KC1 17h PLT CV SFL, n=199, t=-3.4, padj=3.5E-3), and Pam3CSK4 19h (ret Pam3CSK4 19h PLT CV SFL, n=199, t=-3.2, padj=5.9E-3) perturbations were primarily associated with venous thrombosis.

[0104] In addition to diagnostic codes, quantitative lab values that are commonly used to assess various physiological parameters also demonstrated robust associations with bloodperturbation responses. For example, RBC median SSC under 18h LPS condition (ret LPS 18h RBC Med SSC) showed strong positive associations with serum albumin (n=2494, t=l 1.8, padj=3.2E-28) and eGFR (n=2569, t=3.3, padj=5.8E-3), which corresponds with its negative association with CKD status. Significant associations also included clinical traits that are not directly measurable in blood samples, such as a positive correlation between corrected QT interval (QTc) on an electrocardiogram and RBC size variability under 18h LPS perturbation (ret LPS 18h RBC1 SD FSC, n=1946, t=10.2, padj=2.6E-21), indicating that latent whole blood phenotypes can reflect physiological changes occurring in other tissues. In each instance, the molecular basis is accessible through the relevant genetics. In summary, we illustrated that perturbation-based blood readouts are reflective of clinical traits including disease attributions and clinical lab values. Additionally, the inclusion of perturbation conditions elicited previously unknown clinical associations that were not observed at baseline, suggestingperturbations evoked unique previously latent blood cell biological responses that are of disease relevance.

[0105] Multiple chemical stimuli with long exposure durations elicited a distinct population of neutrophils (NE2) in the Sysmex WDF channel, exhibiting high side scatter and fluorescence measurements, which was absent under baseline conditions (see image (b) of FIG. 1 A). As an exemplar, this phenot pe was further investigated and this population of neutrophils was functionally characterized.

[0106] Association of the ratio of NE2 neutrophils to total neutrophil count (NE4) under multiple chemical perturbations with clinical traits showed a significant link between NE2 / NE4 ratio and a complex aggregate of cardiometabolic diseases, specifically atherosclerosis, heart failure, and Type 2 diabetes mellitus (T2DM). For example, the NE2 / NE4 ratio with an inflammatory stimulus (WDF Pam3CSK4 19hNE2 / NE4) had negative associations with T2DM (n=2844, cases=847, t=-5.2, padj=3.8E-06) and obesity' (cases=1112, t=-4.0, padj=4.7E-04) as well as with related lab values (serum triglycerides, n=2248, t=-5.6, p=7.3E-07; serum glucose, n=2657, t=-3.7, padj=1.5E-3). This blood readout also exhibited a positive correlation with total HDL cholesterol levels (n=2259, t=6.3, padj=1.2E-8). These results suggest that alow NE2 / NE4 ratio is associated with cardiometabolic disease phenotypes and relevant clinical lab values.

[0107] Since the NE2 population was only observed with perturbations at later time points, it was hypothesized that it might be related to neutrophil death. To directly evaluate the biological processes occurring in this population, protocols were developed to label purified neutrophils with the Sysmex WDF dye to visualize this population using regular flow cytometry. FIGS. 8A-B illustrate results. The top-left graph of FIG. 8A shows a flow cytometry analysis of isolated neutrophils stained with the Sysmex WDF dye (WDF-APC). NE2-like population is defined as the cells showing elevated SSC and WDF dye intensity. NEl-like population is defined as the main neutrophil population with lower SSC and WDF dye intensity . It was found that the cells that represent the NE2 population, showing elevated WDF dye fluorescence and side scatter, exhibited an increased signal in Annexin V and Sytox, compared to the NEl-like population that mirrors the normal neutrophil profile at baseline. This is further illustrated in the top-center graph of FIG. 8A, which shows the distribution of Sytox-green intensity comparing the NE1- and NE2-like populations, and the top-right graph of FIG. 8A, which shows the distribution of intensities of PE-conjugated Annexin V comparing the NE1- and NE2-like populations. Annexin V is a marker for early apoptosis, while Sytox isindicative of cell death. Furthermore, it was observed that patient blood samples with higher NE2 / NE4 ratio exhibited higher percentages of Sytox and Annexin V positive neutrophils. This is shown in the middle-left graph of FIG. 8A, which shows relationship between Sysmex NE2 / NE4 readout and the percentage of Sytox-green and Annexm V positive neutrophils from blood samples incubated at 39°C for 17h. These results suggest that the unique NE2 population elicited by various chemical perturbations represents a subset of neutrophils actively undergoing apoptosis.

[0108] Recruitment and activation of neutrophils are essential for inflammatory responses. However, delayed apoptosis and impaired clearance of the neutrophils can lead to non-resolving inflammation and subsequent tissue damage. Neutrophils have short lifespans, estimated to range from under 24 hours to 5 days, which can be prolonged by pro-inflammatory and pro-survival signals. Patients with aggregated cardiometabolic diseases exhibited a decreased NE2 / NE4 ratio, suggesting reduced neutrophil apoptosis. It was hypothesized that the reduction in neutrophil apoptosis is a result of their increased pro-inflammatory responses, neutrophil activation and ROS generation were examined at an earlier time point (4.5h posttreatment) that is within the normal range of neutrophil half-life in vivo and compared with the Sysmex NE2 / NE4 readout at a later time point (17h post-treatment). The middle-right schematic of FIG. 8A shows schematics of experimental workflow for comparing neutrophil activation and ROS content at the 4.5h time point and Sysmex readout at the 17h time point for the same blood sample. Neutrophil activation has been previously associated with upregulation of GDI lb on the cell membrane and shedding of CD62L. Using these two surface markers, three distinct neutrophil populations are defined, CDl lbhlghCD62Llow, CDl lbmedlumCD62Lhlg11, and CDllblowCD62Llow. The botom-left graphs of FIG. 8A presents a flow cytometry analysis of isolated neutrophils stained with Alexa 488 conjugated CD62L and Pacific blue conjugated GDI lb from two representative patient samples. The high expression of CD1 lb and shedding of CD62L indicate activated neutrophils, while high surface expression of CD62L suggests quiescent neutrophils and the loss of both surface markers is indicative of cell death.

[0109] A robust anti-correlation between neutrophil activation and NE2 / NE4 was observed, as shown in the botom-right graph of FIG. 8A, which shows the relationship between the percentage of activated neutrophils at 4.5h and the Sysmex NE2 / NE4 readout at 17h for the same donors; as well as between ROS generation (CellROX) and NE2 / NE4 in human subjects, as shown in the top-right graph of FIG. 8B, which shows the relationshipbetween the percentage of activated neutrophils at 4.5h and the Sysmex NE2 / NE4 readout at 17h for the same donors. This suggests that individuals with an increase in proinflammatory neutrophils show a reduced NE2 population in the Sysmex readout. Individual neutrophil trajectories were then tracked with time-lapse imaging using CellROX and Sytox. It was observed that neutrophils that survived until the 15h time point exhibited higher ROS and extended duration with elevated ROS, compared to the neutrophils that died earlier. This can be seen in the graphs of FIG. 8B, which show the time-dependent transient of CellROX deep red (ROS) and Sytox green (Cell death) of neutrophils that survived till 15 h and died before 15 h, and the time of neutrophils become ROS positive comparing cells that survived till 15 h and died before 15h. Together, these results show that neutrophil pro-inflammatory responses, including activation and ROS generation, delay their apoptosis, which is reflected as a reduced NE2 population in Sysmex measurements.

[0110] Note that, in the scatter plots of FIGS. 8A-B, each data point indicates one donor. A linear regression was performed among donors. R2measures the goodness of fit and p value measures the significance of the slope deviating from zero. Each data point indicates an individual neutrophil in L.[OHl] Consistent with the finding that pro-inflammatory neutrophil responses determine the NE2 / NE4 readout, the GWAS results also revealed a SNP rs5743618 in the TLR 1 / 6 / 10 region that is associated with the NE2 / NE4 ratio. FIG. 9 shows that a neutrophil population that emerges in response to inflammatory stimuli and long heat exposure is linked to multiple genetic variants. Differences in flow cytometry measurements among homozygous subjects for multiple genetic loci associated with the response to 19h of Pam3CSK4 exposure at 39C are illustrated. Shown are differences in normalized cell counts in the WDF channel (side scatter and side fluorescence). Significant differences were observed in the distribution of neutrophils for loci near TLR1, CASP3 / ACSL1, PFKP, HK1 and BCL2A1. The rs5743618 variant has been demonstrated to enhance TLR1 trafficking and expression on the plasma membrane. The same SNP has also been shown to account for interindividual variability in Pam3CSK4-induced cytokine responses. To simulate this gam of function in TLR1, the TLR1 / 2 ligand Pam3CSK4 was used. A significant and dose-dependent increase in the NE2 / NE4 profile was found in whole blood treated with Pam3CSK4 when compared to untreated control. FIG. 10 illustrates the neutrophil response to TLR1 / 2 ligand Pam3CSK4.

[0112] The top-left and middle-left graphs of of FIG. 10 show the NE2 / NE4 Ratio fromSysmex readout at 17h in blood incubated with Pam3CSK4 compared to control, and the dose-dependent effect of Pam3CSK4 on NE2 / NE4 ratio at 17h post incubation, respectively. As expected, stimulating neutrophils with Pam3CSK4 also increased the neutrophil activation and ROS generation, as shown in the two top-right, two middle-right, and two bottom-left graphs of FIG. 10. In FIG. 10, the top-right graphs present results of a flow cytometry analysis of isolated neutrophils stained with Alexa 488 conjugated CD62L and Pacific blue conjugated CDl lb under untreated and Pam3 conditions; the middle-right graphs present a histogram of CellRox in neutrophils isolated from blood with or without Pam3CSK4 treatment; and the two bottom-left graphs present the percentage of neutrophil activation under control or Pam3CSK4 conditions, and the percentage of ROS+ neutrophils under control and Pam3CSK4 conditions, respectively. Furthermore, tracking individual neutrophil trajectories revealed that Pam3CSK4 treated cells exhibit prolonged durations of ROS elevation, compared to untreated controls, as shown in the bottom-center-right graph of FIG. 10, which presents the time neutrophils stay ROS+ under control and Pam3CSK4 conditions.

[0113] The bottom-right graph of FIG. 10 shows a seahorse analysis of isolated neutrophils showing ATP produced from the mitochondria and the anaerobic glycolysis pathways in conditions of control and PAM3CSK4 treated. Five replicates were performed for the same donor. Error bars shown indicate SEM of measurements from 4 donors’ blood samples. Treating neutrophils with Pam3CSK4 also greatly increased cell glycolic ATP production (p<0.001 ), suggesting the neutrophils undergo metabolic reprogramming after TLR stimulation, as previously observed in macrophages. These results further support the role of elevated neutrophil proinflammatory responses underlying the increased NE2 / NE4 ratio in the Sysmex experimental system.

[0114] Besides TLR1 and several genes that have been reported to regulate cell death (e.g.n CASP3 and BCL2AP), three additional genes (HK1, PFKP, and ACSLP) involved in cellular metabolism were determined to be associated with the NE2 / NE4 ratio at a genomewide significance. The left schematic of FIG. 11A presents schematics showing the regulatory function of HK1, PFKP, and ACSL1 of neutrophil metabolic pathways. The lead SNPs identified are associated with upregulated expression of HK1 and PFPK, and unknown directionality for ACSL1 (open target database). HK1 and PFKP encode Hexokinase 1 and Phosphofructokinase, two of the key enzymes regulating the rate-limiting steps in glycolysis, converting glucose to pyruvate and generating low levels of ATP. Under anaerobic conditions, pyruvate is further catalyzed to lactate. The lead SNPs identified cause upregulation in HK1 and PFKP expression and decreased NE2 / NE4 ratio, suggesting reduced neutrophil apoptosis.Neutrophils are typically thought to use anaerobic glycolysis as their primary energy source. However, some studies suggest that neutrophils utilize diverse metabolic pathways, including fatty acid oxidation, to provide energy for specific functions. Acyl-CoA synthetase long-chain family member 1, encoded by ACSI.l. converts fatty acid to acyl-CoA, which is then transported into mitochondria for oxidation.

[0115] To investigate the effect HK1 and PFKP modulations in neutrophils, a subsaturation dose of 2-Deoxy-D-glucose (2-DG) (10 mM) was used to inhibit glycolysis and HK1. TheACSZ inhibitor Triacsin C36 was also utilized to study ACSL1 function. First, ATP production from neutrophils was assessed using the Seahorse metabolic analyzer. The top-right graph of FIG. 11 A shows the Seahorse analysis of isolated neutrophils showing ATP produced from the mitochondria and the anaerobic glycolysis pathways in conditions of, water control, 2DG, DMSO control, and Triacsin C. Five replicates were performed for the same donor. Error bars shown indicate SEM of measurements from four donors’ blood samples. Consistent with the literature, it was observed that unstimulated neutrophils are highly glycolytic. As expected, 2-DG decreased glycolic ATP production (p=0.03). In contrast, Triacsin C ablated mitochondrial ATP production (p=0.002) and increased glycolytic ATP (p<0.001). As ACSL1 modulates fatty acid oxidation (FAO), Triacsin C’s effect on FAO was further analyzed using exogenous palmitate as a long chain fatty acid substrate. Compared to the DMSO control, Triacsin C decreased both mitochondrial respiration and maximal respiration in response to FCCP, suggesting a reduced FAO in neutrophils. This can be seen in the bottom-right graph of FIG. 11 A, which shows a Seahorse analysis long-chain fatty acid oxidation rate in neutrophils treated with Triacsin C and DMSO control. Five replicates were performed for the same donor. Error bars shown indicate SEM of measurements from four donors’ blood samples.

[0116] Next, potential mechanisms for how 2-DG and Triacsin C modulate neutrophil function were examined. It was found that both 2DG and Triacsin C increased the NE2 / NE4 ratio in whole blood compared to controls, suggesting an increase in neutrophil death. The topcenter graph of FIG. 11 A shows a Sysmex NE2 / NE4 readout of blood treated with 2-DG, compared to water as control, and Triacsin C, compared to DMSO as control. Measurements were performed after 17-hour incubation at 39°C. Both 2-DG and Triacsin C reduced ROS production in neutrophils. This can be seen in FIG. 11B. The left graphs of FIG. 1 IB show a flow analysis of isolated neutrophils stained with Alexa 488 conjugated CD62L, Pacific blue conjugated GDI lb, and CellRox. Shown are representative dot plot and histogram from one representative sample, comparing water control, 2-DG, DMSO control, and Triacsin Cconditions. The right graphs of FIG. 11 A and FIG. 1 IB show the percentage of activated and ROS+ neutrophils, respectively, at 4.5h post treatment of water control, 2-DG, DMSO control, and Triacsin C. 2-DG also decreased neutrophil activation. Unexpectedly, an upregulation of neutrophil activation induced by Triacsin C was observed. This increase may be caused by the metabolic shift from FAO to glycolysis in neutrophils. The bidirectional effects on neutrophil activation and ROS generation of Triacsin C underlie the smaller effect on neutrophil death observed when compared to 2-DG.

[0117] Lastly, to investigate whether inhibiting HK1 and PFKP using 2-DG promotes neutrophil apoptosis and clearance in vi vo, a transgenic zebrafish model that expresses GFP under the myeloperoxidase (mpo) promoter Tg ( po:GFP) was used to track neutrophil behaviors using fluorescence. Inflammatory responses were stimulated by performing tail transection under control and 2-DG treated conditions. Hyperglycemia was also induced in zebrafish larvae by ablating P-cells. Within 4 hours post tail transection, it was observed that neutrophils were recruited to the injury site, followed by resolution around 24h under the control condition. The left image of FIG. 11C shows a visualization of neutrophils in Tg (mpo.GFP) zebrafish at 4h and 30h post tail transection, comparing control, 2DG, hyperglycemia, and 2DG under hyperglycemia conditions, for four representative zebrafish. The right graph of FIG. 11C shows a quantification of GFP+ cells at the tail transection site at 4h and 30h under control, 2-DG, hyperglycemia, and 2-DG under hyperglycemia conditions. Each data point indicates an individual zebrafish. Adding a sub-saturation dose of 2-DG did not alter the response.

[0118] In contrast, under hyperglycemic conditions, at 24h post tail transection, neutrophils continuously accumulated at the injury site, suggesting a delayed resolution of inflammation. Inhibiting glycolysis pathways through the use of a sub-saturation dose of 2-DG effectively resolved prolonged neutrophil inflammation at the injury' site under the hyperglycemia condition. These ex vivo and in vivo results suggest that HK1, PFKP, and ACSL1 interact to regulate neutrophil inflammatory responses in vivo by modulating their metabolic profiles. Pharmacological inhibition of HK1 and PFKP is effective in preventing sustained inflammation related to hyperglycemia and promoting neutrophil clearance. SNPs leading to upregulation of HK1 and PFKP expression reduce the NE2 / NE4 ratio, which is prevalent in cardiometabolic disease. Patients with these common alleles appear to exhibit delayed inflammation resolution, contributing to disease pathologies. Thus, modulation of HK1 and PFKP may potentially serve as a mechanism-driven therapeutic strategy for such patients.

[0119] As correlations between blood-response readouts and clinical traits were observed, tests were performed to determine whether polygenic scores (PGSs) based on blood response summary statistics can be used to stratify patient populations and improve predictions of clinical events. PGSs were calculated for perturbation-blood responses spanning different cell types and conditions using clumping and thresholding for subjects in the MassGeneral Brigham (MGB) Biobank, a large cohort of subjects with available genotypic data and comprehensive electronic health records. Cox proportional hazard models were computed for 20 clinical outcomes using blood-based PGSs for 327 readouts, adjusting for sex and the first two genetic principal components. The PGSs calculated from each blood readout exhibited unique associations with specific diseases. FIGS. 12A-F present the set of PGSs calculated from perturbation-based blood responses, and shows that they are associated with differences in time to onset of diseases. FIGS. 13A-J present an overview of all PGS-disease associations. FIGS. 12A-D collectively show a matrix of hazard ratio estimates in time-to-event models for clinical outcomes in the MGB Biobank using polygenic scores of blood readouts under different perturbation conditions. Time to first diagnostic code or diagnosis date in medical problem list was modeled using sex, first two genetic principal components and scaled bloodresponse scores. Points indicate significant associations after multiple testing correction across all tested diseases and blood traits (20 clinical outcomes and 327 blood readouts) with adjusted p-value thresholds: • 0.05, • • 0.01, • • • 0.001.

[0120] From FIGS. 12A-E and 13A-J, significant associations can be identified for multiple common cardiac / cardiometabolic conditions; for example, obesity (pltf Colchicine 20h IPF Med FSC, cox padj=4.31E-4, cases=8985), atherosclerosis (pltf Isobutyric 3h IPF CV FSC, cox padj=4.85E-4, cases=6860), chronic kidney disease (ret KC1 17h RBC CV SFL, cox padj=2.86E-5, cases=5538), and heart failure (ret KC1 17h RET2 SD SSC, cox padj=2.2E-3, cases=4290). In addition, strong associations were also observed with rare immune system conditions such as juvenile arthritis (wnr Nigericin 0.5h PLT SD SSC, padj=0.026, cases=89) and glomerular disease in systemic lupus erythematosus (ret Colchicine 20h PLT CV SFL, cox padj=0.0056, cases=64).

[0121] In addition to proportional hazards models, subjects were stratified into quartiles according to their PGS and plotted the time to the first diagnosis for a subset of diseases and blood traits, which showed clear separation among different quartiles. This can be seen in detail in FIG. 12D, which presents survival curves for a subset of diseases stratified by blood-response PGSs. P-values are based on a test-for-trend after grouping subjects accordingto their PGS quartile. For example, the 4th quartile based on PGS calculated from variability in RET2 SSC under 17h KCL perturbation (ret KCL 17h RET2 SD SSC) showed delayed onset of heart failure compared to the first three quantiles, suggesting the genetic basis underlying this blood cell trait can be used to predict risk for heart failure and explore the mechanisms leading to heart failure development.

[0122] Next, the direction of association of polygenic models in the MGB Biobank was compared with blood readouts measured in the screening cohort. Out of 47 blood trait PRS- disease pairs with padj < 0.01 in the MGB Biobank, 31 blood readout-disease pairs had at least 20 cases with available perturbation response measurements, 16 of which were significantly associated in both cohorts and had consistent directionality of association, and 15 were not significantly associated in the screening cohort at padj < 0.01.

[0123] Blood readouts associated with variants inACSLl, PFKP. wA HK! were further investigated in detail. As described above, these metabolic genes regulate neutrophil activation and clearance, particularly under hyperglycemia. Thus, readouts were analyzed to test whether PGSs calculated based on these blood cell traits predict the time to onset and progression of CKD in subjects with prediabetes and diabetes (hbAlC>5.7). CKD stages 3a, 3b, 4 and 5 were categorized, based on eGFR thresholds (<6000, <45, <30, and <15 mL / mm / 1.73m2). It was found that the PGSs for RBC variability in SFL under 21h Alhydrogel, 20h Colchicine, 17h KCL, and 18h LPS perturbations were positively associated with the progression of CKD. In comparison, the NE2 / NE4 ratio under 17h KCL, 20h Colchicine, and 19h Pam3CSK4 conditions was negatively associated with the development of CKD. FIG. 12E presents hazard ratio estimates for the progression to different CKD stages in subjects with pre-diabetes and diabetes using PGS models of blood traits that had significant associations with ACSL1, PFKP o HKl. Points indicate significant associations after multiple testing correction with adjusted p- value thresholds: • 0.05, • • 0.01, • • • 0.001. These results suggest that PGSs based on cellular readouts can be used to identify subpopulations of diabetics and prediabetics at increased risk of discrete complications, in this example accelerated progression of CKD.

[0124] Over 3300 traits have been investigated using GWAS in more than one million participants, with current studies continuing to increase the sample size to improve the statistical power. While the techniques are robust, and hundreds of thousands of genetic loci have been found to be associated with diseases, it remains difficult using comparative methodologies to identify those loci with consequential biological effects. One major bottleneck in GWAS using the comparative methodologies is a generalizable strategy to movefrom a locus to mechanistic insights and identified genetic target, limiting the translation toward therapeutic development. As set forth above, the present disclosure outlines an approach that combines cellular phenotyping with GWAS, to uncover previously latent large effect size genetic loci with direct implications for cell biology. Using multigenic models based on selected cellular phenotypes, clinical phenotypes with significantly altered disease risks related to these intermediate phenotypes were identified.

[0125] The initial focus was on cellular responses in peripheral blood, as such samples are highly accessible, have long been used as a diagnostic tool in clinical settings, and the technologies are broadly available for subsequent scaling of any useful findings. More recently, GWAS have identified loci determining individual variations in blood cell parameters, providing insights into how genetic background influences hematopoietic disorders. In addition to clinically available assays of cross-sectional cellular counts, blood cell properties were assessed under 36 chemical and biological perturbation conditions, aiming to elicit cellular phenotypes that are latent at baseline, and thus likely to be previously unmeasured. This approach was chosen to favor the identification of previously unknown disease-related intermediate phenotypes, from which it was possible select those associated with large effect size common alleles. This method could enable the identification of Mendelian or pseudo- Mendelian loci that might represent rigorous human drug targets. Using this perturbational method, the phenotypic space was expanded from 29 blood parameters used in previous studies to over 4000 blood cell readouts. It was possible to identify alleles associated with important cellular processes, such as neutrophil activation and apoptosis, which play roles in common complex diseases beyond hematopoietic disorders. Evoked cellular response traits in peripheral blood offer a complementary approach to existing phenotyping with the potential to identify genes and pathways with high translational and clinical relevance.

[0126] To validate that risk genes identified using our framework are truly linked to disease-relevant cell biology, a functional study was conducted of genes associated with the evoked NE2 population in the Sysmex readout showing an elevated WDF dye intensity. Although the Sysmex measurements are not tailored to characterize neutrophil function, the WDF dye (a nucleic acid dye) used to distinguish blood cell lineages was found to be reflective of neutrophil apoptosis, possibly as a result of changes in membrane integrity that result in an altered dye uptake. The delay in neutrophil apoptosis was due to a neutrophil pro-inflammatory response. The evoked response assays enabled the efficient identification and in vitro and in vivo experimental validation of genes involved in the metabolic HK1,PFKP, w ACSLl) andinflammatory pathways (TLR1 / 6 / 1O) that affect neutrophils and their lifespan. It was demonstrated that a shift in neutrophil metabolic profile results in a prolonged increase in intracellular ROS, which then delays neutrophil apoptosis. The findings illustrated an association between common human genetic variants in metabolic genes and neutrophil innate immune responses, revealing cell metabolism as a potential therapeutic target for inflammation in various cardiometabolic diseases at a broader population level.

[0127] The perturbational blood cell-based approach described herein also reveals common genetic variants with large effect sizes. Notably, several genes that were identified have been demonstrated to be causative Mendelian genes for specific diseases. For example, common coding variants in TUBB1 that affect platelet traits were identified by the methodologies herein, while rare variants in TUBB1 were identified to cause inherited thrombocytopenia. BMPR2, which is linked to hereditary pulmonary arterial hypertension (PAH), was found to be associated with monocyte responses in this study. As monocytes and macrophage abnormalities have been identified as part of the pathophysiology of PAH, this finding suggests a monocytic contribution to the vascular inflammation observed in BMPR2- linked PAH, but also offers a direct window into potential somatic contributions to other forms of PAH. These examples support the utility of latent phenotypes to define previously unknown cellular mechanisms that can bridge common genetic variations and complex diseases.

[0128] Polygenic scores calculated from blood cell traits associated with these metabolic genes showed utility in risk prediction for renal complications of diabetes. An emerging body of evidence supports the involvement of innate immunity in CKD initiation and progression in diabetes, but comparative studies have typically focused on macrophages. The results set forth herein reveal a role for genetically determined variation in the genesis of pro- inflammatory, long-lived neutrophils in the development of CKD in patients with prediabetes or diabetes. The blood readout-based polygenic score models were able to stratify patients with distinct risks in developing various cardiometabolic, vascular, and inflammation diseases, revealing patient cohorts that might benefit from therapeutics targeting related biological pathways. These findings suggest testable therapeutic strategies for targeting pro-inflammatory neutrophils in the prevention and treatment of CKD in subsets of T2DM.

[0129] In support of the efficacy of the methodologies set forth herein, blood cell phenotyping was performed using an available cytometry device that is primarily designed for robust whole blood cell counts. These methodologies may be extended by, for example, implementing standardized evoked responses across multiple phenotyping tools, such asquantifying functional dyes with flow cytometry and cell morphology using imaging, which may expand the accessible phenotypic space. The framework presented herein that incorporates human genetic data, primary cellular phenotyping, and deep clinical traits, enables reiterating of genetic risk loci discovery, systematic target validation, and subsequent drug discovery. Implementation of such a method in routine clinical care will facilitate the development of refined clinical trajectories, the annotation of common genetic variation, and the identification of large effect size heritable contributions from such alleles to human disease.

[0130] Experiment and Statistical Analysis Details

[0131] Physiologically relevant doses and time points were determined for each perturbation to elicit reproducible effects on blood analyzed on a SysmexXN-1000 hematology analyzer. Table 1 above presents perturbation condition descriptions including dose and exposure times. Compounds dissolved in DMSO or Chloroform were prepared such that the percent by volume of solvent is < 0.5%. Each condition was assigned a three-digit identifier (e.g., -007) which was paired with a patient ID for each treated sample (e.g., AA-00100-007). This standardized label scheme allowed for the preparation of barcoded sample tubes and batch-wise automated measurements using the hematology analyzers.

[0132] Up to 40 patients per day were recruited from multiple phlebotomy clinics and donated up to 50ml of blood in addition to their clinical blood draw. Whole blood was collected in 8.5mL ACD tubes (BD 364606). Barcoded sample tubes with patient and perturbation identifiers were aligned and prepared batch-wise, by aliquoting 700 pL of whole blood into a grid of 5mL round bottom tubes. All perturbation compounds were added to blood at specific time points and transferred to incubator shakers (39C, 200 RPM). After incubation, tubes were placed in automated sampling racks and profiled using the Sysmex XN-1000. Both Sysmex- derived blood parameters (e.g., CBC) as well as raw cytometry data were exported as .csv and .fcs files.

[0133] Before aliquoting patient blood samples, a portion of freshly drawn blood was set aside to be used in whole blood DNA extraction. DNA was extracted from 3mL of whole blood using Qiagen Puregene Blood Core Kit C (158389). DNA was quantified and checked for quality using NanoDrop One and Qubit, diluted to 75 ng / pL, and stored at -80 °C. Samples were aliquoted into 96-well barcoded plates and quantified using Cytation Take3 Trio before genotyping. Internal genotyping for qualify control was performed using Adv anta Sample ID Genotyping Panel (Fluidigm 101-7773). Aliquots were shipped to Northwell Health Genomics Alliance and University of Miami Genotyping Core in 96-well barcoded plates with one emptywell for controls. Samples were quantified using Nanodrop and Qubit to identify plates with high numbers of low concentration samples, which could be replaced before genotyping. Genotypes were called from genomic DNA in batches of approximately 500 samples using the Illumina GSAv3 Beadchip and Illumina Genome Studio.

[0134] Samples were excluded from participants that had high missingness of variants (>10%), sex mismatches from genotyped data, or had withdrawn from the study. In addition, for samples failing Advanta fingerprinting (concordance of at least 0.75 in at least 20 SNPs), genotyping was repeated, or the samples were removed. Variants with high missingness across subjects (>10%) or deviations from Hardy-Weinberg equilibrium at a p-value <le-50 were filtered. Structural or multi-allelic variants were removed. A local instance of Michigan Imputation Server vl.4.1 with Eagle2 and Minimac4 was used to impute genotypes with the 1000G Phase3 v5 reference panel. Post imputation, variants with minor allele frequency <0.0001 were removed. The first 10 principal components were estimated using plink2. Relatedness was estimated using plink2 with the KING-robust kinship estimator and 5 subjects with a kinship greater than 0. 177 (first-degree relations or closer) were removed. In total, after these exclusions, genotype data were available for 2685 subjects on >3.5 million imputed variants.

[0135] MGB samples were genotyped in batches using three related Illumina arrays (MEGA, MEGA Ex and MEG), as well as the Illumina GSAv3 array. Imputation was performed for using Michigan Imputation Server with the 1000G Phase3 v5 reference panel for each batch. Batches were merged using the intersection of variants present in all batches and applied the same QC filtering as above. In short, subjects with high missingness (>10%) or sex mismatches were removed. Variants with high missingness across subjects (>10%) or deviations from Hardy-Weinberg equilibrium at a p-value < le-50 were filtered. Structural or multi-allelic variants were removed. PC A was calculated using plink2 and subjects with kinship greater than 0.177 as well as subjects with non-European ancestry (distance greater than 3x radius of 1000G EUR reference samples in joint PCA) were removed using plinkQC. Subjects that were part of the screening cohort were removed from the MGB Biobank cohort. In total, after these exclusions, genotype data were available for 44705 participants on >6.7 million imputed variants. For polygenic score applications, variants were further filtered to have a minimum allele count of 100 and missingness of less than 2% leaving 1.8 million variants.

[0136] A total of 330 blood-based cellular phenotypes were measured using a blood flow cytometer (Sysmex XN-1000) under 36 different conditions. The blood cell parameters can be categorized into indices related to membrane / intracellular structure measured using side scatter (SSC), nucleic acid and membrane lipid content measured using side fluorescence (SFL), and cell shape / volume measured using forward scatter (FSC), as well as parameters such as cell counts and percentages within defined regions (gates). For each parameter, robust estimators such as median, robust standard deviation, and robust coefficient of variation were calculated using FlowJo vl0.8. Gates were empirically defined based on densities of measured cells under baseline and perturbation conditions and included additional regions for subpopulations that were typically not observed under baseline conditions. A total of 15 whiteblood cell-related gates, 7 RBC gates, 4 platelet-related gates were defined as well as 4 gates for debris or unknown cell types. All samples were measured within 36 hours of blood draw, with baseline measurements occurring within 3h for 80% and 7h for 95% of samples.

[0137] Thorough quality control was performed to identify sources of technical variation as well as biological covariates. For this, several factors were assessed including the effect of the time between blood draw and flow cytometry measurement, drift over the course of the study (study month), as well as biological covariates such as age, sex, and race (see FIGS. 4A-B). Outlier samples (i.e., samples where a single phenotype was outside of 4 median absolute deviations from the median measurement of all samples under the same conditions) were removed. Additionally, a 2-dimensional PCA projection was computed for all blood measurements from a single fluorophore under a single perturbation condition and samples that were further than 2 median absolute deviations from the median sample were removed. Finally, the phenotypic measurements were quantile-transformed. The final numbers of blood measurements as well as genotyped subjects passing QC across conditions are shown in Table 2, above.

[0138] Flow cytometry analyses were performed on neutrophils isolated from patients’ whole blood samples, using the Easy Sep Direct Human Neutrophil Isolation Kit (#19666, STEMCELL). Post isolation, neutrophils were resuspended in Tyrode’s solution as described previously. To characterize the NE2-like cell population using flow cytometry, neutrophils were isolated from whole blood samples that were incubated at 37°C for 17h, and then labeled with apoptosis indicators, Sytox green (S7020, ThermoFisher Scientific), and R-PE conjugated annexm V (ThermoFisher Scientific). The labeled neutrophils were then subjected topermeabilization using Sysmex WDF Lysercell (Sysmex) and staining with Fluorocell WDF dye (Sysmex). The samples were analyzed 5 minutes post the addition of Flurocell WDF dye.

[0139] To characterize neutrophil activation and ROS, isolated neutrophils were labeled with Pacific Blue anti-human GDI lb antibody (Biolegend, Clone ICRF44) and Alexa Fluor 488 anti-human CD62L antibody (Biolegend, Clone DREG-56). Cells were then subsequent! ally labeled with CellROX Deep Red Reagent (ThermoFisher Scientific, Cl 0422) at 37°C for 30 minutes. Cells were washed and resuspended in staining buffer prior to flow cytometry analyses.

[0140] For the Real-Time ATP Rate Assay, a DMEM assay medium containing 1 OmM glucose, ImM pyruvate, and 2mM glutamine was used. Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were measured from neutrophils isolated from patients’ whole blood pretreated with or without 2DG (10 mM) or Triacsin C (5pg / mL), using a Seahorse XFe96 Analyzer. Neutrophils were resuspended in DMEM medium and seeded (IxlO6per well) in a Seahorse 96 well plate coated with CellTak (Coming, 354240) for 20 minutes. Cell attachment was visually confirmed prior to the assay. The assay was performed according to manufacture instructions. 1.5 pM Oligomycin, 1 pM FCCP, and 0.5 pM Rotenone / Antimycin A were used.

[0141] For the long-chain fatty acid stress test, neutrophils isolated from untreated whole blood were first resuspended and incubated for 2h at 37°C in a substrate-limited medium containing 0.5mM glucose, ImM glutamine, 0.5mM L-Camitine, and 1% FBS. Cells were then pelleted and resuspended in an assay medium containing 2 mM glucose, and 0.5mM L- Camitine. Cells were and seeded (1x106per well) in a Seahorse 96 well plate coated with CellTak (Coming, 354240) for 20 minutes. After visually confirming cell attachment, cells were treated with Triacsin C (5pg / mL) or DMSO control for 30 minutes. Palmitate-BSA fatty acid oxidation substrate was added prior to the assay. The assay was performed according to the manufacturer’s instructions. 4 pM Etomoxir, 1.5 pM Oligomycin, 1 pM FCCP, and 0.5 pM Rotenone / Antimycin A were used. Normalization for both assays was performed based on direct cell counting.

[0142] Zebrafish larvae at 54h post-fertilization were anesthetized by immersion in E3 with 4.2% tricaine. Tail transections were performed with a sterile scalpel at the distal end of the notochord. Bnghtfield and fluorescence images were acquired with a Cytation 5 at 4h and 30 h post transection at 28°C. A neutrophils count within the tail region was performed using ImageJ.

[0143] After genetic and phenotypic QC, blood phenotypes were retained for 4723 subjects and genotypes for 2685 subjects. Debris, ghost, and unknown cell type gates were excluded from genetic association tests because they yielded non-normally distributed phenotypes after quantile transformation. A univariable GWAS was performed for each of the remaining 286 traits under 36 different conditions. The cohort consisted primarily of subjects with primarily European ancestry, preventing robust multi-ancestry analyses due to low numbers of subjects in other ancestry groups. Therefore, associations were calculated and reported for the subset of participants with European ancestry only. Specifically, plink2 was used to compute association statistics for a linear regression of phenotype on the allele dose for >3.5 million imputed variants with minor allele frequency >0.05, minor allele count >10, covariate variance standardization and the covariates age, sex, time from blood draw to analysis, month of study, genotyping chip and batch, and the first 10 genotype principal components.

[0144] A threshold of 5E-8 was used as a significance threshold for each phenotype and no corrections were performed for multiple testing at the level of association p-values. Many of the measured phenotypes were correlated across similar gate / cell types (e.g., subpopulations of neutrophils), phenotypic dimensions (e.g., SSC and FSC), or conditions (e.g., TLR ligands Pam3CSK4 and LPS). Given the large number of tests and limited number of subjects, a concise set of variants was sought that are associated with the strongest observed cellular responses. For this all significant variants were clumped using plink, requiring LD >0.50, physical distance <250kb between clumped variants, and at least two independent hits from different traits for each clumped region. The variant with the smallest association p- value across all measured traits for a given region was used as the lead variant. Further associated regions were annotated using gene range lists provided by plink2 (https: / / www.cog- genomics.org / static / bm / plink / glist-hgl9). Each region was also annotated with associations based on a comparative analysis by Vuckovic et al.

[0145] Twenty binary clinical phenotypes were defined using ICD10 diagnostic codes (see Table 4), 17 of which had more than 20 cases in the screening cohort. Twenty-six quantitative measurements available across the entire cohort were collected, such as the comprehensive metabolic panel, lipid panel, and structured electrocardiographic data. Linear or logistic models were fitted, associating binary and continuous traits with 327 blood phenotypes that showed significant genetic associations to at least one locus. These modelsalso included the covanates age, sex, race, and time from blood draw to measurement. FDR- adjusted q-values were calculated using the R package qvalue.

[0146] For 327 traits with significant genetic associations, summary statistics from the screening cohort were used to calculate polygenic scores for subjects in the MassGeneralBrigham Biobank. The clumping and thresholding approach was used to calculate polygenic scores. Specifically, command plink -clump-pl 0.5 -clump-r2 0.5 -clump-kb 100 was used for clumping and a p-value threshold of 0.1 was used for the scoring step. The same disease definitions as above were used to define case status, as well as the age at first diagnostic code or first addition to the medical problem list as event date. Cases with diagnoses specified in their medical history without a specific first diagnosis date were excluded from the time-to- onset analyses. Then, Cox proportional hazards models were calculated for the time to onset of 20 clinical outcomes with the variables sex, first two genetic principal components and polygenic score for 327 blood traits. In addition, subjects were stratified into quartiles and used to the log-rank test for a subset of PGS-disease associations, using only the blood-trait polygenic scores and no additional covanates.

[0147] In addition to the blood flow cytometer (Sysmex XN-1000)-based platforms, as noted above, the present disclosure may be implemented on any tool for characterization of peripheral blood cells, such as a flow-cytometry-based platform to assess blood lineagespecific functional response to pharmacological perturbations. In combination with an annotated compound library, this platform allows one to determine patient- and cell lineagespecific ion channel activities and other proteins’ functions (e.g., gene, RNA, protein, metabolite, and / or biological pathway functions). Using these data, together with patient’s electronic health records and genetic data, it is possible to build disease association models to identify novel biomarkers and therapeutic targets for various diseases.

[0148] In one example, the workflow of the flow cytometr -based analysis of peripheral blood samples is as follows. The whole blood samples collected are subject to a magnetic or other physical / chemical procedure of cell enrichment or separation (in this example, magnetic RBC depletion) first to enable efficient cell labeling and reduce the time needed for flow cytometry analysis of each sample. In other examples, the procedure may be or include chemical red blood cell depletion or a cell enrichment. Processed cells are labeled with fluorescence functional indicators, together with RNA / DNA dyes, to gate for specific cell lineage. The labeled cells are then distributed into 96 or 384 well plates and chemical orpharmaceutical perturbations are added. The plates containing samples are then subject to high- throughput flow cytometry analysis.

[0149] To process the flow cytometry data, gates were designed based on FSC, SSC, and DNA / RNA dye intensity to separate cell types. FIG. 14 illustrates an example lineagespecific response to Yodal using the high throughput flow cytometry method. In the top row of FIG. 14, a gating strategy to isolate different blood cell types based on FSC, SSC, and Hoechst is shown. Using this method, it is possible to quantify patient-specific Piezol responses in RBCs, neutrophils, lymphocytes, platelets, and eosinophils with Piezol -specific activator Yodal. This is illustrated in the bottom row of FIG. 14, which shows the Ca2+response (X-Rhodl) in different lineages under control or Yodal conditions.

[0150] FIG. 15 illustrates another example using this method, in which TRPA1 (Polygoidal), NaV1.3 (ICA121431), BKCa (NS11021), and TRPM4 (CBA) activities in each blood cell type were assessed across 500 patients’ blood samples. FIG. 15 shows the most affected blood cell parameters for each ion channel-targeting perturbation. DPBS1 is the control condition. The results may then be analyzed with machine learning methods, such as a decision tree, to determine the primary cell feature for each perturbation. FIGS. 16A-B illustrate an example decision tree analysis to identify primary cellular readouts for each ion channel targeting perturbation of FIG. 15. Using this platform, an annotated target compound was screened and several targets were identified that modulate blood cell functional readouts. These results are illustrated in Table 5.Table 5 - Identified Compounds and Targets for Modulating Blood Cell Function

[0151] Example Implementations

[0152] In examples, the methodology set forth herein may be implemented as a method of profiling blood cell traits. The methods set forth herein may be used to analyze blood sample and genetic data from a subject (e.g., a human subject) to identify previously unknown disease biomarkers or drug targets, and / or to define subsets of diseases to achieve precise, mechanistic- driven pharmacological intervention. FIG. 17 illustrates one example of such a method 1700. The method 1700 includes an operation 1702 of evoking a latent blood cellular phenotype, including adding at least one of a chemical compound or a biological perturbation to the blood a blood sample. The blood sample may be a human peripheral blood sample, such as whole blood or enriched / isolated blood cell samples.

[0153] The method 1700 further includes an operation 1704 of analyzing the evoked blood cellular phenotypes (i.e., the phenotypes measured or determined by operation 1702) using one or more assays. The assays may include any combination of flow-cytometry assays, microscopy / imaging-based assays, hematological / blood counter-based assays (e.g., Sysmex assays), and the like. Operation 1704 may include detecting at least one of SSC, FSC, or FLS of the cellular population, for example using a fluorescent dye.

[0154] At operation 1706, a disease association model is developed or generated, thereby to output a biomarker for an associated disease or disease subset. The disease association model may be based on the assayed blood cellular phenotype (i.e., as measured or determined by operation 1704) and a disease outcome record. The disease outcome record maybe or include at least one of an electronic health record, a diagnostic code record, or lab measurements. In examples, the disease outcome record may be implemented as a database. In examples, the disease association model may implement a statistical method (such as a linear or logistic regression) and / or a machine learning method. Accordingly, operation 1706 identifies the evoked blood phenotypes that may serve as previously unknown biomarkers for associated disease or previously undefined disease subsets.

[0155] Next, the method 1700 includes an operation 1708 of identifying a therapeutic compound to modulate the biomarker (i.e., the biomarker determined by operation 1706). Operation 1708 may include applying at least one assay (e.g., of the type described above with regard to operation 1704) to perform chemical screens on blood cells to identify previously unknown therapeutic compounds (or known compounds that were not previously known to have a therapeutic effect with regard to a particular disease or subset) that modulate the biomarker. In examples, operation 1708 may be followed by an operation 1710 of applying the identified therapeutic compound to treat the identified disease or disease subset and / or causing the application of the identified therapeutic compound (e.g., generating and transmitting instructions to cause the application, initiating the treatment, etc.). Additionally or alternatively, operation 1710 may include generating a report, which may be written or digitally displayed, to inform further clinical practice. The report may indicate the identified therapeutic compound and / or further information on the anticipated or projected modulation of the biomarker, to thereby clinical decision making, which would otherwise not be possible without the preceding steps that are performed by a computer system, computer processor, and / or blood assay analysis system.

[0156] FIG. 18 illustrates another example of a method 1800 for profiling blood cell traits. The method 1800 includes an operation 1802 of evoking a latent blood cellular phenotype, including adding at least one of a chemical compound or a biological perturbation to the blood a blood sample. The blood sample may be a human peripheral blood sample, such as whole blood or enriched / isolated blood cell samples.

[0157] The method 1800 further includes an operation 1804 of analyzing the evoked blood cellular phenotypes (i.e., the phenotypes measured or determined by operation 1802) using one or more assays. The assays may include any combination of flow-cytometry assays, microscopy / imaging-based assays, hematological / blood counter-based assays (e.g., Sysmex assays), and the like. Operation 1804 may include detecting at least one of SSC, FSC, or FLS of the cellular population, for example using a fluorescent dye.

[0158] At operation 1806, a GWAS may be performed using the evoked blood cell phenotypes (e.g., as measured or determined by operation 1804), thereby to determine or identify a druggable target. The druggable target may be a previously unknown druggable target (e.g., a target that was previously unknown, or a known target that was not previously known to be druggable) for an associated disease or disease subset.

[0159] Next, the method 1800 includes an operation 1808 of identifying a therapeutic compound to modulate the druggable target (i.e., the druggable target determined by operation 1806). Operation 1808 may include applying at least one assay (e.g., of the type described above with regard to operation 1804) to perform chemical screens on blood cells to identify previously unknown therapeutic compounds (or known compounds that were not previously known to have a therapeutic effect with regard to a particular disease or subset) that modulate the biomarker. In examples, operation 1808 may be followed by an operation 1810 of applying the identified therapeutic compound to treat the identified disease or disease subset, causing initiation of the application or treatment, and / or generating a report, as described.

[0160] Operations 1808 and / or 1808 may result in the identification of previously unknown compositions of matter. For example, as set forth above, three compounds (SP 100030, ML 365, and (R)-CR8) were found to selectively modulate an evoked neutrophil trait (NE2 / NE4 in the Sysmex array under KC1), which is indicative of inflammation and is associated with the progression of chronic kidney disease in diabetic populations. These compounds show potential as therapeutic agents for resolving inflammation in this population. In some implementations, the methods 1700 and 1800 may be combined, such that both operation 1706 and operation 1806 are performed in either order.

[0161] The whole blood perturbation screening may include a determination of physiological relevant doses and time points for perturbations. After perturbation, the assaying may include an analysis of blood using a hematology flow cytometry analyzer or general flow cytometer on a blood sample (e.g., whole blood; RBC partially depleted blood labeled with a functional dye such as mitotracker, sytol6, fluo4, etc.). In examples, the samples may be present as a preparation of compounds dissolved in DMSO or Choloroform, ensuring a specific concentration. A standardized three-digit identifier may be assigned for each condition and paired with a subject ID to create unique identifiers for treated samples.

[0162] FIG. 19 illustrates one example of a blood analysis system 1900 that may be used to implement the methodologies set forth herein, including the methods 1700 and / or 1800. The blood analysis system 1900 includes at least one processor 1902, at least one memory1904, and an input / output (I / O) interface 1906. The blood analysis system 1900 may be configured with various modules (e.g., various software modules) to implement blood analysis functions. In an example, the modules may be present in a non-transitory computer-readable medium (e.g., the memory 1904) in the form of instructions that, when executed by the processor 1902, cause the blood analysis system 1900 to perform any one or more of the operations described herein. In another example, the processor 1902 may be configured to load and / or execute instructions from another non-transitory computer-readable medium (e.g., cloud storage or from the memory of another device). The blood analysis system 1900 may also include an assay system 1908. The assay system 1908 may include one or more of an optical, microscopy, flow cytometry, or hematologic analyzer configured to receive and analyze a sample as described above. For example, the assay system may be configured to perform a screen on blood cells, feedback from which is communicated to the processor 1902 to identify a therapeutic compound that modulates a desired biomarker.

[0163] The processor 1902 may include one or more individual electronic processors, each of which may include one or more processing cores, and / or one or more programmable hardware elements. The processor may be or include any type of electronic processing device, including but not limited to central processing units (CPUs), graphics processing units (GPUs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, digital signal processors (DSPs), or other devices capable of executing software instructions. When a device is referred to as “including a processor,” one or all of the individual electronic processors may be external to the device (e.g., to implement cloud or distributed computing). In implementations where a device has multiple processors and / or multiple processing cores, individual operations described herein may be performed by any one or more of the microprocessors or processing cores, in series or parallel, in any combination. In some implementations, one or more of the processing units or processing cores may be remote (e.g., cloud-based).

[0164] The memory 1904 may be any storage medium, including a non-volatile medium, e.g., a magnetic media or hard disk, optical storage, or flash memory; a volatile medium, such as system memory, e.g., random access memory (RAM) such as dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), extended data out (EDO) DRAM, extreme data rate dynamic (XDR) RAM, double data rate (DDR) SDRAM, etc.; on-chip memory; and / or an installation medium where appropriate, such as software media, e.g., a CD-ROM, or floppy disks, on which programs may be stored and / or datacommunications may be buffered. The term “memory” may also include other types of memory or combinations thereof. For the avoidance of doubt, cloud storage is contemplated in the definition of memory. A memory is an example of a non-transitory computer-readable medium which stores instructions that are executable by a processor (or processors), the execution of which causes the executing device (e.g., a computer) to perform certain operations, such as those operations described herein.

[0165] The I / O interface 1906 may include interface components to permit the communication of data to and from external devices or sources. For example, the I / O interface 1906 may include communication ports and / or interfaces to permit communication with other computer devices. The communication ports and / or interfaces may permit input and output via wired protocols (e.g., Ethernet, Universal Serial Bus (USB), FireWire, etc.) and / or wireless protocols (e.g., Wi-Fi, Bluetooth, Near Field Communication (NFC), 5G, 4G, etc.). The I / O interface 1906 may additionally or alternatively include communication ports and / or interfaces to permit communication with a user. For example, the I / O interface 1906 may include interfaces for a mouse, a keyboard, a display, a graphical user interface (GUI), buttons, switches, etc.

[0166] Methods 1700 or 1800 may further include operations such as preparation of barcoded sample tubes, calibration of the cytometer or other assay (e.g., using standards on a daily basis), automated batch-wise measurements using the cytometry analyzers or other analyzers, and the like. New QC lots may be regularly acquired (e.g., every 28 days) following manufacturer guidelines. Whole blood may be collected in specific tubes, such as ACD, EDTA, or heparin tubes. The samples may further be subjected to batch-wise preparation and aliquoting of whole blood into a specified grid of tubes. Perturbation compounds may be added to blood at specific time points, and the samples may then be incubated under specific conditions (e.g., 20°C, 37°C, 39°C, 200 RPM, etc.). Samples may be automatedly profiled using the cytometer or other assay.

[0167] Cellular parameters may be defined, and phenotypes may be subject to quality control. These operations may include categorization of blood cell populations and parameters into various types based on the characteristics, such as membrane / intracellular structure, nucleic acid and membrane lipid content, and cell shape / volume. Empirical gate definitions of cell types may be established based on densities of measured cells under various conditions. The operations may further include a calculation of robust estimators such as median, robust standard deviation, and robust coefficient of variation for each gated cell population. Thequality control assessment may consider factors such as time between blood draw and measurement, study month drift, and biological covariates like age, sex, and race. Outliers may be detected and removed based on the criterion of deviation from median measurements for each individual readout. These operations may include computation of a projection of sample parameters using dimension reduction algorithms such as PCA, t-SNE, UMAP, for blood measurements under specific conditions and outlier removal based on distance from other samples under the same condition. The operations may include quantile transformation or other normalization of the phenotypic measurements.

[0168] In addition to phenotypes, genotype quality control may be performed. This may include operations such as exclusion of samples with high missingness of variants, sex mismatches, or those withdrawn from the study; filtering out variants with high missingness or deviations from Hardy -Weinberg equilibrium; genotype imputation using reference panel such as 1000G Phase3 v5 and methods such as Eagle2 and Mimmac4; filtering out post-imputation variants with very low minor allele frequencies; and / or Principal Component and Kinship Analysis using methods such as plink2 and exclusion of subjects with high kinship values.

[0169] The GWAS may be univariate GWS, and may be performed for all readouts under all conditions with methods such as Plink2 or Regenie, using linear models or wholegenome regression models. Variant clumping may be performed, in which significant variants are grouped (or “clumped”) using plink with certain criteria (e.g., LD r2 >0.50, physical distance <250kb). The variant with the smallest association p-value across all traits in a region may be chosen as the lead variant. Clumped regions may be annotated with nearby candidate genes based on genomic location. When multiple genes are associated with a location, a prioritization system can be used by prioritizing coding variants, intronic variants, and distance to the transcription start site (TSS).

[0170] In associating phenotype readouts with clinical endpoints, the clinical phenotypes may be defined using structured electronic health record data such as ICD10 diagnostic codes or lab values such as comprehensive metabolic panel, lipid panel, or structured electrocardiographic data. Quantile transform or other normalization operations may be performed. The association may also be based on a fit of logistic or linear or other statistical models associating clinical outcomes with blood phenotypes that demonstrate significant genetic associations. Incorporate covariates into the model, including age, sex, race, and time from blood draw. Models may be tested using statistical tests such as z-test and t-test, and may be adjusted for multiple testing.

[0171] In associating polygenic scores with clinical endpoints, summary statistics from the screening cohort may be used to define polygenic scores in a separate cohort using methods for polygenic score calculation such as clumping and thresholding or lassosum. Clinical endpoints may be defined in the same way as for phenotype associations above, such as using structured HER data such as diagnostic codes, lab values, or ECG measurements. One may define time-to-event (TTE) models using time to first observed diagnosis post-birth and use methods such as Cox’s proportional hazards regression for fitting TTE models, including variables such as sex, the genetic principal components, and polygenic score for blood traits, may fit linear models or logistic regression models between polygenic score for blood traits and continuous or categorical clinical outcomes; and / or may use appropriate statistical tests to determine significant associations between blood-based polygenic score and clinical outcome and adjust for multiple testing.

[0172] One example of TTE modeling of blood readouts with clinical endpoints may be performed as follows. First, clinical phenotypes are defined using structured electronic health record data, for example primarily focusing on ICD10 diagnostic codes to identify specific conditions or diseases. Next, blood readouts (perturbation response values) are standardized using z-score normalization to create risk scores. Survival analysis is then performed using Cox proportional hazards models to associate clinical outcomes with blood phenotypes. In this, the primary outcome may be time to event, calculated as the duration between the blood readout collection date and either the date of first diagnosis for cases or the last contact date for controls. Cases are defined as individuals who received the relevant ICD10 diagnostic code during the follow-up period. Cases with a first diagnosis before blood draw or within a 6-month time period of the blood draw are excluded. Controls are individuals who did not receive the diagnostic code by their last recorded contact date. Cox models incorporating covariates such as sex, age at blood draw, and risk score (e.g., as a continuous variable) may be used. Model performance is evaluated using hazard ratios and their confidence intervals for each variable; p-values for each variable, particularly focusing on the risk score; concordance statistics to assess model discrimination; or combinations thereof.

[0173] The above example TTE analysis was performed in an experimental analysis, and revealed significant associations between different blood cell perturbation response values and later diagnoses of several cardiometabolic diseases, including CKD, cardiomyopathy, and T2D with nephropathy. For CKD, Cox proportional hazards model showed a significant association with multiple risk scores based on monocyte perturbation responses(wdf]Pam3CSK4 19h|MO_Count: log HR = 0.38, SE: 0.06, p = 1.72e-6, wd^LPS 18h|MO_Count: log HR = 0.35, SE: 0.05, p = 2.13e-6). The models demonstrated good discriminative ability with a concordance statistic of 0.74-0.75. Cardiomyopathy showed a significant association with a risk score derived from RBC side fluorescence (ret|KCl 17h|RBC_CV_SFL: log HR = 0.42, SE: 0.08, p = 0.00053). The model for cardiomyopathy had a concordance statistic of 0.742, indicating good predictive performance. For type 2 diabetes with nephropathy, we observed a significant association with a risk score based on WBC count (pltf|Pam3CSK4 19h|WBCl_Count: log HR = 0.49, SE: 0.1, p = 0.00064). The model showed strong discriminative ability with a concordance statistic of 0.75. FIG. 20 illustrates Kaplan-Meier curves for selected associations.

[0174] Across these diagnoses, the models adjusted for age and sex had significant associations between the risk scores and disease outcomes. The case counts varied across conditions, with 148 cases for CKD, 75 cases for cardiomyopathy, and 50 cases for T2D with nephropathy. These findings suggest that blood cell perturbation responses may serve as early indicators of cardiometabolic disease risk and could lead to mechanistic insights into disease progression.

[0175] The systems and methods set forth herein, as evidenced by the results presented above, establish the principles required to fully individualize the target ID, drug discovery and development process for a broad range of diseases where the relevant latent phenotypes can be defined. The overall platform approach not only brings together care and discovery while linking the two fields directly with previously unknown computable biological metrics, it also enables: ab initio identification of human target and screening assays; immediate and ongoing human relevance throughout the discovery / development process-including lead optimization, toxicology studies, target population refinement, target engagement assays, clinical trial entry / endpoints (companion diagnostic); avoidance of the issues with many other Al-based in vitro screening systems including the lower representativeness of in vitro culture, lack of complexity or excess complexity (e.g., iPS differentiation protocols) and the burden and cost of very extensive animal modeling; the potential for matching disease and drug mechanism of action at an individual subject level.

[0176] Other examples and uses of the disclosed technology will be apparent to those having ordinary skill in the art upon consideration of the specification and practice of the invention disclosed herein. The specification and examples given should be consideredexemplary only, and it is contemplated that the appended claims will cover any other such embodiments or modifications as fall within the true scope of the invention.

[0177] The Abstract accompanying this specification is provided to enable the United States Patent and Trademark Office and the public generally to determine quickly from a cursory inspection the nature and gist of the technical disclosure and in no way intended for defining, determining, or limiting the present invention or any of its embodiments.

Claims

CLAIMSWhat is claimed is:

1. A method of analyzing a blood sample, comprising: evoking a latent blood cellular phenotype, including adding at least one of a chemical compound or a biological perturbation to the blood sample; assaying the evoked blood cellular phenotype with at least one assay system; generating a disease association model based on the assayed blood cellular phenotype and on a disease outcome record, thereby to output a biomarker for an associated disease or disease subset; and using the at least one assay system, performing a screen on blood cells to identify a therapeutic compound that modulates the biomarker.

2. The method of claim 1, wherein the blood sample is a whole blood sample or an enriched blood cell sample.

3. The method of claim 1 or claim 2, wherein the blood cells are neutrophils.

4. The method of any one of claims 1 to 3, wherein the at least one assay system includes at least one of an optical, microscopy, flow cytometry, or hematologic analyzer.

5. The method of any one of claims 1 to 4, wherein assaying the evoked blood cellular phenotype includes detecting at least one of a side scatter, a forward scatter, or a fluorescence scatter of the cellular population using a fluorescent dye.

6. The method of any one of claims 1 to 5, wherein the disease outcome record includes at least one of electronic health record data, diagnostic codes, or lab measurements.

7. The method of any one of claims 1 to 6, wherein the disease association model includes a statistical model.

8. The method of claim 7, wherein the statistical model is a linear regression model or a logarithmic regression model.

9. The method of any one of claims 1 to 6, wherein the disease association model includes a machine learning model.

10. The method of claim 9, wherein the machine learning model is a decision tree model.

11. The method of any one of claims 1 to 10, wherein evoking the latent blood cellular phenotype includes incubating the blood sample after adding the at least one of the chemical compound or the biological perturbation.

12. The method of any one of claims 1 to 11, wherein evoking the latent blood cellular phenotype includes performing a magnetic or chemical red blood cell depletion or a cell enrichment on the blood sample before adding the at least one of the chemical compound or the biological perturbation.

13. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor of a blood analysis system, cause the system to perform operations comprising the method of any one of claims 1 to 12.

14. A method of analyzing a blood sample, comprising: evoking a latent blood cellular phenotype, including adding at least one of a chemical compound or a biological perturbation to the blood sample; assaying the evoked blood cellular phenotype with at least one assay; performing a genome wide association study (GW AS) on the evoked blood cellular phenotype, thereby to generate a druggable target for the associated disease or disease subset; using the at least one assay, performing a screen on blood cells to identify a therapeutic compound that modulates the druggable target.

15. The method of claim 14, wherein the blood sample is a whole blood sample or an enriched blood cell sample.

16. The method of claim 14 or claim 15, wherein the blood cells are neutrophils.

17. The method of any one of claims 14 to 16, wherein the at least one assay includes at least one of an optical, microscopy, flow cytometry, or hematologic analyzer.

18. The method of any one of claims 14 to 17, wherein assaying the evoked blood cellular phenotype includes detecting at least one of a side scatter, a forward scatter, or a fluorescence scatter of the cellular population using a fluorescent dye.

19. The method of any one of claims 14 to 18, wherein the disease outcome record includes at least one of electronic health record data, diagnostic codes, or lab measurements.

20. The method of any one of claims 14 to 19, wherein the disease association model includes a statistical model.

21. The method of claim 20, wherein the statistical model is a linear regression model or a logarithmic regression model.

22. The method of any one of claims 14 to 19, wherein the disease association model includes a machine learning model.

23. The method of claim 22, wherein the machine learning model is a decision tree model.

24. The method of any one of claims 14 to 23, wherein evoking the latent blood cellular phenotype includes incubating the blood sample after adding the at least one of the chemical compound or the biological perturbation.

25. The method of any one of claims 14 to 24, wherein evoking the latent blood cellular phenotype includes performing a magnetic or chemical red blood cell depletion or a cell enrichment on the blood sample before adding the at least one of the chemical compound or the biological perturbation.

26. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor of a blood analysis system, cause the system to perform operations comprising the method of any one of claims 14 to 25.

Images