System and method for target or biomarker identification

Abstract

A method for identifying key targets or biomarkers of a disease at a cell specific level according to a two-step process that comprises: (1) identifying cell type-specific candidate targets / biomarkers with genetically defined causal relationships to one or more disease features by performing causal analysis (e.g., 2 sample mendelian randomization); and (2) ranking candidate targets / biomarkers by analyzing the interconnectivity between the identified candidate targets / biomarkers and common interactors / regulators in a customized cell type-specific knowledge graph.

Description

[0001] SYSTEM AND METHOD FOR TARGET OR BIOMARKER IDENTIFICATION

[0002] TECHNICAL FIELD

[0003] The present disclosure is directed to computer-implemented methods of identifying a target or biomarker for a disease. Particularly, but not exclusively, the present disclosure is directed to identifying a target or biomarker for a disease using causal analysis. Particularly, but not exclusively, the present disclosure is directed to identifying a target or biomarker for a disease using causal analysis and knowledge graphs.

[0004] BACKGROUND

[0005] Identifying cell type-specific contributions to a specific disease is difficult, especially as populations of patients can be highly heterogeneous due to differences in, for example, genetic contribution to disease, symptom presentation, or disease progression trajectory.

[0006] Many groups have attempted to resolve cell type-specific genetic contributions to disease (or other phenotypes) but fail to evaluate causal relationships, such as between cell types and disease features, and / or rely on predictions of cell type.

[0007] One approach to resolving cell type-specific genetic contributions to disease is to use a single cell type quantitative trait loci (QTL) as an exposure variable for analysis, by methods such as by mendelian randomization. Dang et al., Mov. Disord., 2022; 37(12):2451-2456, for example, ran two sample mendelian randomization (2SMR) analyses using dopaminergic neuron-specific expression quantitative trait loci (eQTL) to nominate causal genes in Parkinson's Disease.

[0008] Some studies have contextualized 2SMR results (generated using bulk -omics data as an exposure variable) in a secondary cell type-specific dataset. For example, McGowan et al., Human Molecular Genetics, 28, 2019, 3293-3300, conducted 2SMR on circulating cytokines and immune-mediated diseases and subsequently looked at gene expression from three immune cell type datasets (i.e., monocytes, neutrophils, and T cells) to see if any identified causal associations could be attributed to a single cell type.

[0009] Other studies deconvolute data generated using bulk -omics data, to predict cell typespecific contribution(s). For example, Westra et al., PLoS Genet 11(5): el005223 (2015), conducted a gene-environment interaction meta-analysis to infer cell type specific contribution derived from whole blood eQTL.

[0010] Multiple tissue-level exposure variables have been evaluated in relation to disease / trait outcome(s) using 2SMR analyses in a number of studies including Richardson et al., Nat. Commun. 11, 185 (2020), who conducted 2SMR on 48 tissue-specific eQTL and 395 complex traits; and Taylor et a / ., Genome Med. 11, 6 (2019), who conducted 2SMR on 44 tissue-specific eQTL and 14 cardiovascular-related traits. Porcu et al., Nat. Commun. 10, 3300 (2019), developed a transcriptome-wide summary statistics-based MR approach, which uses multiple SNPs as instruments and multiple gene expression traits as exposures simultaneously, and applied this method to 48 tissue types and 43 traits.

[0011] Cell specificity and function in health and disease is determined by interactions between different genes or proteins at the level of transcription, translation, and post-translational modifications. Because of the high level of interconnectivity / interdependency between genes, a disease rarely arises due to a single gene function defect but usually reflects larger perturbations on interconnected genes that otherwise are normal. This rationale, together with the large accumulation of biological knowledge, has made appealing the use of knowledge-graphs or biological networks for the discovery of novel disease genes or modules, and therefore targets. For example, Jensen et al., Circulation: Cardiovascular Genetics. 2011;4:549-556, identified novel susceptibility genes for coronary artery disease by analyzing neighboring genes to significant genome-wide association studies (GWAS) hits via a protein-protein interaction network. Vlietstra et al., PLoS ONE 17(7): e0271395 (2022), identified genes targeted by disease-associated non-coding SNPs using a protein knowledge graph. Himmelstein et al., PLoS Comput. Biol., 11(7): 61004259. (2015) and Binder et al., Commun. Biol., 5, 125 (2022), used knowledge-graphs for the prioritization and prediction of novel disease-associated genes, respectively.

[0012] The identification of disease genes in a cell type-specific manner has become possible with the development of single cell (scRNAseq) and single nuclei (snRNAseq) RNAseq technologies. However, the development of computational methods involving these types of datasets has been restricted to individual (intra)dataset group comparisons based on modern statistical methods, without considering data integration or expansion to include biological networks.

[0013] Many experimental and / or computational methods exist to parse cell type-specific contributions to a disease phenotype, including single cell gene expression profiling (e.g., Human Cell Atlas), multiplexed imaging / spatial transcriptomics (e.g., Allen Brain Atlas - Aging, Dementia, and TBI Study), or cellular deconvolution, e.g., cellular fractions in AML samples.

[0014] SUMMARY OF DISCLOSURE

[0015] According to a first aspect of the present disclosure there is provided a method of identifying key targets or biomarkers for a disease. The method comprises obtaining, by one or more processors, data for genetic risk variants associated with a first feature of the disease, obtaining, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature, and determining, by the one or more processors, one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features. The method further comprises identifying, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features and generating, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes is represented by one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors. The method further comprises calculating, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of Interconnectivity is identified as a key target or biomarker.

[0016] The identification of a key target enables a therapy to be provided to treat or prevent the disease, wherein the therapy is targeted to the key target, or in some embodiments to a biological pathway that is associated with the key target. Biological pathways associated with a key target are biological pathways that are directly affected by modification of the key target and propagate the effects of modifying the key target. Accordingly, targeting another part of the biological pathway can also be used to treat or prevent the disease.

[0017] The identification of a key biomarker enables detection of the biomarker to enable diagnosis of the disease. The identification of a key biomarker can be used to stratify a patient population so that a more personalized therapeutic treatment can be provided.

[0018] One aspect of the method of the present disclosure is to resolve genetic contributions to disease at a cell type-specific level to improve target identification for drug discovery programs, as known genetic drug mechanism of actions are associated with increased likelihood of drug approval (see Nelson et al., Nat. Genet., 2015, 47(8):856-60).. The method of the present disclosure provides cell type resolution of targets and their relationship to specific disease feature(s) - incorporating specific, identifiable patient groups or subgroups directly into the target discovery process. The method of the present disclosure enables the generation of a matrix of results defining causal relationships between cell types and disease features, thereby allowing prioritization of which cell type to further evaluate using cell type-specific knowledge graphs. The method can also accelerate downstream drug optimization processes, including biology prioritization, assay development, patient selection / stratification, and biomarker identification.

[0019] The method of the present disclosure uses cell type-specific knowledge graphs for the contextualization of candidate targets or makers to derive additional biological insights and to identify broader biological changes in pathways and / or functions. This is a significant advantage over analyses focusing on single genes or proteins as being the primary putative targets / biomarkers.

[0020] A gene or protein associated with a genetic risk variant may or may not be targeted therapeutically, or may not be desirable to target therapeutically (e.g., due to known side effects or known difficulties in generating drugs specific for the target). If a gene or protein associated with a genetic risk variant is expressed in multiple cell types, the method of the present disclosure enables insights into which cell type should be prioritized, e.g., based on incidence and / or effect size of causal relationships between said cell type and a disease feature. If a gene or protein associated with a genetic risk variant cannot be targeted therapeutically, the method of the present disclosure can be used to identify and rank druggable genes or proteins associated with genetically anchored disease biology while maintaining cell type specificity within the curated knowledge graph.

[0021] According to a second aspect, the present disclosure provides a method of treating a subject suffering from a disease, wherein a key target identified by the method according to the first aspect of the present disclosure is targeted by a therapy in order to treat or prevent the disease. In certain situations, a plurality of lead targets may be identified and one or more of the plurality of lead targets may be targeted by therapy in order to treat or prevent the disease. The therapy can be any suitable therapy ( / .e. , any suitable therapeutic modality) depending on the nature of the target. If the target is a gene sequence, gene therapy can be used to knockout or reduce expression of the gene. If the target is a protein, a molecule capable of interacting with the protein to block or reduce its function, e.g., an antibody molecule or a small molecule, can be used. In some embodiments, the key target is not a known target for the disease, i.e., the target was not previously disclosed as being a target for the disease. The key target may be a known i target for a different disease, or it may not be known as a target for any disease. The treatment can be provided in vivo or ex-vivo.

[0022] According to a third aspect, the present disclosure provides a method for determining whether a subject has a disease, wherein the method comprises determining the presence of a key biomarker in a sample obtained from the subject, wherein the key biomarker is identified by the method according to the first aspect of the present disclosure. In situations where a plurality of key biomarkers have been identified, the method may comprise determining the presence of the plurality of key biomarkers in the sample in order to determine whetherthe subject has the disease. Depending on the key biomarkers, and the specific cell type for which the causal relationship between the biomarker and the disease has been determined, different methods and samples can be used to detect the presence of the biomarkers.

[0023] According to a fourth aspect, the present disclosure provides a system comprising one or more processors and a memory storing instructions which, when executed by the one or more processors, cause the one or more processors to perform the method according to the first aspect of the present disclosure.

[0024] According to an additional aspect of the present disclosure, there is provided a method of identifying key targets or biomarkers for a disease. The method comprises obtaining, by one or more processors, data for genetic risk variants associated with a plurality of features of the disease, obtaining, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a plurality of cell types, wherein each of the molQTL is associated with a molecular feature, and determining, by the one or more processors, a matrix of genetic variants having a causal relationship with one or more of the plurality of features of the disease and one or more associated molecular features for the plurality of cell types. The method further comprises identifying, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features and generating, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors. The method further comprises calculating, by the one or more processors, a level of interconnectivity for each node within the first cell typespecific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker. Additional aspects and embodiments of the present systems are disclosed, and the above aspects and embodiments should not be construed as limiting the present disclosure.

[0025] In accordance with the above, and with the disclosure herein, the present disclosure includes improvements in computer functionality or in improvements to other technologies at least because the disclosure herein discloses a process for creating and storing a cell type-specific knowledge graph which is used to identify a key target or biomarker. The process for creating and storing such a knowledge graph, when deployed on the underlying system, allows the systems and methods of the present disclosure to execute with fewer iterations, and use fewer computing resources, than prior art related systems and methods. That is, the present disclosure describes improvements in the functioning of the computer itself of "any other technology or technical field" because the cell type-specific knowledge graph of the present disclosure allows the underlying computer system to utilize less processing and memory resources compared to prior art systems and methods. This is at least because the cell type-specific knowledge graph provides a compact and informative representation of the cellular network related to a candidate target or marker without the need for further experimentation, analysis, or computer simulations across a wide range of tests using multiple compute cycles and data. Therefore, use of the cell type-specific knowledge graphs of the present disclosure results in fewer compute cycles, or otherwise iterations, that has less of an impact on the underlying computing device compared to previous prior art systems and methods that do not use the compact and informative representation of the cellular network. Said another way, the systems and methods of the present disclosure improve over the prior art at least because prior art systems and methods require an empirical or trial-and-error approach that can involve real-world trials regarding identification of key targets or biomarkers that can result in, and require, large database and memory utilization and processor usage to arrive at a similar real-world or simulated results that has a same or similar result. On the other hand, the disclosed systems and methods describe use of a compact and informative representation of the cellular network comprising a cell type-specific knowledge graph, which is implemented for candidate target or marker identification within the graph itself, and without the need for further experimentation, which requires less memory usage and / or processing utilization compared to a conventional approach where large sets of unknown, potentially irrelevant data is used or required. Moreover, the disclosure herein allows for identification and use of high-fidelity data sets, as part of the cell type-specific knowledge graph, which reduces the need for additional computational cycles of the underlying computing platform. i Further, the present disclosure effecting a transformation or reduction of a particular article to a different state or thing, e.g., the transformation or reduction of a cell typespecific knowledge graph to a graphical representation thereof. The graphical representation of a cell type-specific knowledge graph helps to improve the identification of key targets or biomarkers, or common interactors, by providing visual feedback on a graphical user interface (GUI) or otherwise display (e.g., a display screen) regarding relationships of target(s) or biomarker(s), including key target(s) or biomarker(s), within the knowledge graph which may not be identifiable or otherwise inferred from a nongraphical representation. Moreover, the graphical representations generated make efficient use of screen space thereby reducing the display requirements (e.g., display screen sizing requirements) whilst also advantageously reducing the visual influence of nodes which are unlikely to be identified as being a key target or biomarker.

[0026] Still further, the present disclosure includes specific features other than what is well- understood, routine, conventional activity in the field, and / or otherwise adds unconventional steps that confine the disclosure to a particular useful application, e.g., systems and methods for identifying key targets or biomarkers for a disease, generating cell type-specific knowledge graphs, and / or identifying a key target or biomarker, which, for example, can be used for performing drug discovery and development.

[0027] Advantages will become more apparent to those of ordinary skill in the art from the following description of the preferred embodiments which have been shown and described by way of illustration. As will be realized, the present embodiments may be capable of other and different embodiments, and their details are capable of modification in various respects. Accordingly, the drawings and description are to be regarded as illustrative in nature and not as restrictive.

[0028] BRIEF DESCRIPTION OF DRAWINGS

[0029] Embodiments of the present disclosure will now be described, by way of example only, and with reference to the accompanying drawings, in which:

[0030] Figure 1 shows a method of identifying key targets or biomarkers for a disease according to an aspect of the present disclosure;

[0031] Figure 2 shows steps which may be performed with the method of Figure 1 according to an embodiment of the present disclosure;

[0032] Figure 3 illustrates 2-sample mendelian randomization; i Figure 4 shows a matrix of causal relationships according to embodiments of the present disclosure;

[0033] Figure 5A shows a knowledge graph according to an embodiment of the present disclosure;

[0034] Figure 5B shows a cell type-specific knowledge graph generated from the knowledge graph shown in Figure 6A according to an embodiment of the present disclosure;

[0035] Figures 6A-6D show visualizations of matrices of 2-sample mendelian randomization results according to an example of the present disclosure;

[0036] Figure 7 shows a knowledge graph associated with the matrices shown in Figures 6A-6C according to an example of the present disclosure;

[0037] Figure 8 shows a knowledge graph associated with matrices shown in Figures 6A-6D according to an example of the present disclosure; and

[0038] Figure 9 shows an example computing system for carrying out the methods of the present disclosure.

[0039] DETAILED DESCRIPTION

[0040] The ability to provide a cell type-specific, directionally informed, process for identifying candidate targets / biomarkers with causal links to a disease feature (e.g., a clinical feature of Parkinson's disease) enables key targets / biomarkers to be identified with a high degree of confidence. To provide such a process, the present disclosure is directed to the integration of molecular features within a wider biological context using causal analysis and knowledge graphs. Cell type-specific candidate targets / biomarkers with genetically defined causal relationships to one or more disease features are identified by performing causal analysis (e.g., mendelian randomization). Then, candidate targets / biomarkers are ranked by analyzing the interconnectivity between the identified candidate targets / biomarkers and common interactors / regulators in a customized cell type-specific knowledge graph. Advantageously, the present disclosure contextualizes the causally identified candidate targets / marked within a wider biological context to allow broader biological changes in pathways and / or molecular function to be identified (rather than focusing solely on single gene / protein candidate targets). The key target(s) and / or biomarker(s) may be used, for example, for performing drug discovery and development.

[0041] Figure 1 shows a method 100 of identifying key targets or biomarkers for a disease according to an aspect of the present disclosure. The method 100 comprises the steps of obtaining genetic risk variant data (step 102), obtaining cell type-specific molecular quantitative trait loci (molQTL) data (step 104), determining genetic risk variants having a causal relationship with one or more molecular features (step 106), identifying candidate targets or biomarkers (step 108), generating a cell type-specific knowledge graph (step 110), and calculating levels of interconnectivity (step 112). The method 100 also comprises the optional steps of identifying a key target or biomarker, or common interactor (step 114), and outputting the key target or biomarker, or common interactor (step 116). The method (100) may also comprise identifying one or more biological pathways associated with the key target or key targets. Biological pathways associated with a key target(s) are biological pathways that are directly affected by modification of the key target(s).

[0042] In general, the method 100 identifies and prioritizes cell type-specific targets / biomarkers relevant for treating or diagnosing a disease. More particularly, genetically defined causal relationships between candidate targets / biomarkers, specific cell types, and disease features are determined so as to determine whether candidate targets / biomarkers and corresponding disease features are unique or conserved across cell types and / or disease features. The cell type-derived candidate targets / biomarkers for the disease indication are then contextualized within a customized cell type-specific knowledge graph (KG) as nodes. In doing so, the original candidate targets / biomarkers are expanded to include other targets / biomarkers (e.g., genes or proteins) identified as common interactors / regulators connecting thereto. This helps to enrich the pool of cell type-specific candidate targets / biomarkers (e.g., by 5-10 times). In addition, enriched biological pathways can be identified within these expanded networks. Given the sparse nature of candidate targets / biomarkers by cell type with a significant causal link, this step improves the ability to interpret target / biomarker function and / or significance within a broader cellular network context and better inform target / biomarker identification with confidence. This in turn enables prioritization for targets / biomarkers or pathways. Critically, by applying graphbased analytics on the cell specific KG data, candidate targets / biomarkers can be analyzed based on their interconnectivity / centrality and allow the identification of key (i.e., strong) targets / biomarkers with greater confidence. Overall, the method provides an iterative and modular framework for target / biomarker identification.

[0043] At the step 102, data for genetic risk variants is obtained. The data for genetic risk variants is associated with a first feature of the disease. In some embodiments, data for genetic risk variants associated with a plurality of features of the disease is obtained.

[0044] A "genetic risk variant" refers to any genetic variation in a nucleic acid sequence that is associated with one or more features of the disease. In certain embodiments, the genetic risk variant may be a single-nucleotide polymorphism / variation (SNP / SNV), an insertion and / or deletion variation, a copy number variation, a translocation and / or an inversion, or a combination thereof. In particular embodiments, the genetic risk variant is a SNP. In certain embodiments, the genetic risk variant is identified from genome-wide association studies (GWAS) that associate the genetic risk variant with the one or more features of the disease. In certain embodiments, the "genetic risk variant" is any genetic variation in a nucleic acid sequence that is associated with one or more features of the disease for a particular subgroup of patients, e.g., a subgroup of patients identified by biomarkers (e.g., genetic markers) and / or a comorbidities.

[0045] A "feature of the disease" refers to a phenotypic characteristic of the disease and, in some embodiments, includes disease incidence, i.e., presence of the disease. A disease is characterized by a set of features, e.g., symptoms, morphological changes, functional changes, etc. Many diseases share common features; however, the combination of the features results in a disease diagnosis. In the present disclosure, genetic risk variants can be associated with the presence of the disease and / or associated with a feature of the disease, i.e., one feature from a plurality of features that are used to diagnose the disease. In certain embodiments, the feature of the disease refers to a phenotypic characteristic of a disease that on its own does not enable the disease to be definitively diagnosed. For example, for Parkinson's disease, "a features of the disease" may refer to motor score severity, age of onset, cognitive impairment, depression, daytime sleepiness, dyskinesia, hyposmia, insomnia, RBD, MMSE score and Hoehn and Yahr scale. In certain embodiments, the term "feature of a disease" only refers to a phenotypic characteristic of the disease and does not include disease incidence, i.e., presence of the disease.

[0046] Within the context of the present disclosure, the term "disease" refers to any disease, including heart disease, such as ischemic heart disease and coronary artery disease; cancers; bacterial and viral infections; cerebrovascular diseases, such as stroke; respiratory diseases, including chronic obstructive pulmonary disease; diabetes; autoimmune diseases, such as ulcerative colitis, Crohn's disease, inflammatory bowel disease, rheumatoid arthritis, Guillain-Barre syndrome, Sjogren's syndrome, scleroderma and Graves' disease; and neurological disorders, including Alzheimer's disease, multiple sclerosis, Parkinson's disease, amyotrophic lateral sclerosis (ALS) and frontotemporal dementia. In some embodiments, the disease is Parkinson's disease.

[0047] The data for genetic risk variants can be obtained from any suitable source, including GWAS summary statistics. Example databases of GWAS summary statistics include those listed in Table I. TABLE I

[0048] In one implementation, the GWAS data is pre-processed so that data for a genetic risk variant is associated with one or more features of the disease and the association meets a statistically significant threshold— i.e., a relationship or association between two or more features are statistically significant. Any suitable method can be used to determine the statistical significance, as known to those skilled in the art. The threshold can be set at the level of standard level of significance, e.g., p < 0.05, or can be set so that the level is of greater statistical significance, e.g., p < 0.01, p < 0.001, or p < 0.0001. In a certain embodiment, the statistically significant threshold, p, is p < 0.0001. The skilled person will appreciate that such pre-processing may be performed using any suitable software such as the ieugwasr package for the R programming language, or vcf2gwas, PyGWAS, or Hail for the Python programming language.

[0049] At step 104, cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type is obtained. Each of the molQTL is associated with a molecular feature. In certain embodiments, cell type-specific molecular quantitative trait loci (molQTL) data for a plurality of cell types is obtained.

[0050] Molecular quantitative trait loci (molQTL) refers to any genetic variant that is associated with a molecular feature, such as level of expression (eQTL), level of DNA methylation (meQTL), histone modification (hQTL), chromatin accessibility (caQTL), alternative gene splicing (sQTL), protein levels pQTL), microRNA expression (mirQTL) and ribosome occupancy (rQTL). See Vandiedonck, Clin. Genet., 93, 520-532, 2018. The molQTL data as used herein is cell type-specific molQTL data. In other words, the molQTL data is for a specific cell type. In certain embodiments, the genetic variant associated with the molecular feature is a single single-nucleotide polymorphism / variation (SNP / SNV), an insertion and / or deletion variation, a copy number variation, a translocation and / or an inversion, or a combination thereof. In particular embodiments, the genetic variant is a SNP.

[0051] In certain embodiments, the molQTL data is cell specific eQTL data. The genetic variant associated with the level of expression can be a single single-nucleotide polymorphism / variation (SNP / SNV), an insertion and / or deletion variation, a copy number variation, a translocation and / or an inversion, or a combination thereof. In particular embodiments, the molQTL data is cell specific eQTL data and the genetic variant is a SNP associated with the level of expression.

[0052] The cell type-specific molQTL data may be obtained from one or more sources such as Uffelmann et al., Biological Psychiatry, 2021, 89:41-53; the DICE Database; BIOSQTL; BloodeQTL; EQTLcatalogue; EQTLGen; GTEx; scRNA_eQTLsj. BRAINEAC; CMC;. PsychENCODE; and xQTLServer.

[0053] In a similar manner to the genetic variant data described above, the molQTL data may be pre-processed so that data for the specific cell type is obtained such that the association between the molQTL data and the molecular feature meets a statistically significant threshold (as defined above).

[0054] As used throughout the present disclosure, "a specific cell type" is any defined cell type that is phenotypically and / or functionally distinct. The term may refer to a cell type, cell subtype or cell phenotype. For example, the term may refer to a cell lineage, such as T cells, B cells, natural killer cells, monocytes, etc.; or to a cell subtype, such as T-cell subtypes (e.g., CD4 naive, CD4 stimulated, CD8 naive, CD8 stimulated, TFH, TH1, TH2, TH1 / 17, TH17, Treg naive, and Treg memory cells); or a cell phenotype, such as Ml and M2 macrophages.

[0055] In certain embodiments, the specific cell type is not a general class of cells, such as immune cells, blood cells or muscle cells, but specific cell types and / or specific cell subtypes.

[0056] The use of cell type-specific molecular quantitative trait loci (molQTL) data enables accurate causal effects to be identified at a cell type-specific level. For example, if molecular quantitative trait loci (molQTL) data from whole blood is used (instead of using cell type-specific data), for a target that is causally associated with multiple cell types, an average of its effect size will be seen, which will be driven by the most robust effect, and can hide differences in magnitude and / or directionality of that association. For example, in PD incidence 2SMR and KANSL1 had positive and negative causal associations, respectively, with respect to the risk of developing PD depending on cell type.

[0057] At step 106, one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features are identified.

[0058] This step enables the identification of genetic risk variants having a causal relationship with the disease feature and one or more associated molecular features. In certain embodiments, the identification of the causal relationship enables the identification of directionality between the genetic risk variants and the disease feature.

[0059] One or more genetic risk variants having a causal relationship with the disease feature and one or more associated molecular features are determined by integrating the data for the one or more genetic risk variants with the cell type-specific molQTL data using a causal analysis process to determine the causal relationship. In one embodiment, the causal relationship is determined using a mendelian randomization approach such as 2 sample mendelian randomization (2SMR), which is explained in more detail in relation to Figure 3 below.

[0060] By performing causal analysis (e.g., 2SMR) a causal relationship between one or more genetic risk variants and the one or more disease features at a cell specific level can be identified. The causal relationship meets a statistically significant threshold, e.g., p < 0.0001. In embodiments using mendelian randomization, the statistical significance of the causal relationship is determined as part of performing the mendelian randomization analysis. Suitable methods include those described in Richardson et al., 2020 (supra); Taylor et al., 2019 (supra); Porcu et al., 2019 (supra), and the TwoSampleMR R module. In certain embodiments, the TwoSampleMR R module is used to perform the mendelian randomization analysis. In certain embodiments, the use of mendelian randomization analysis enables the identification of directionality between the genetic risk variants and the disease feature.

[0061] The identification of the causal relationship between the one or more genetic risk variants and the one or more disease features at a cell specific level is based on the identification of a genetic risk variant that is associated with the disease feature and associated with the molecular quantitative trait. In embodiments, the data for the one or more genetic risk variants and the cell typespecific molQTL data are harmonized and / or filtered prior to performing causal analysis. Harmonizing the data ensures that the genetic risk variant data and the cell type-specific molQTL data are comparable and that shared genetic features, e.g., identical SNPs, in the two data sets can be identified. Harmonizing the data can be achieved by mapping the data to a single reference genome. Methods for performing such harmonization include any suitable genome coordinate conversion, or remapping tool, such as CrossMap or the Bioconductor rtracklayer package for the R programming language.

[0062] The step of filtering the data prior to performing causal analysis comprises identifying shared genetic features, e.g., identical SNPs, in the two data sets. Genetic features that are not shared can be filtered out, i.e., removed from the data sets prior to integrating the matched data for the one or more genetic risk variants with the matched cell typespecific molQTL data using causal analysis. Such filtering steps can be easily performed using any suitable software package, e.g., TwoSampleMR.

[0063] In embodiments where genetic risk variant data is obtained for a plurality of different disease features at step 102 and / or molQTL data is obtained for a plurality of different cell types at step 104, step 106 includes generating a matrix of genetic variants. The genetic variants have a causal relationship with one or more of the features of the disease and one or more associated molecular features for the cell types.

[0064] As described in more detail in relation to Figure 4 below, the matrix of genetic risk variants comprises a table of values quantifying the causal relationship that cell type-molecular feature combinations have on a disease feature outcome. As such, the matrix of genetic risk variants may be generated by estimating the causal relationship (e.g., calculating the causal effect using mendelian randomization) for each combination of genetic variant, disease feature, and cell type-specific molecular feature of the data obtained. For example, a matrix may be generated for a specific disease feature and quantify the causal relationship between genetic variants and a plurality of specific cell types (see Figure 4). Alternatively, a matrix may be generated for a specific cell type and quantify the causal relationship between the genetic variants and a plurality of different disease features. A 3D matrix may also be generated to quantify the causal relationship between a plurality of a specific disease features, the genetic variants and a plurality of specific cell types.

[0065] At the step 108, a plurality of candidate targets or biomarkers of the disease are identified based on the one or more associated molecular features. Furthermore, in embodiments where a matrix of genetic risk variants has been generated, step 108 may comprise i identifying the candidate targets or biomarkers of the disease from the matrix of genetic risk variants (as described in more detail in relation to Figure 4 below).

[0066] Candidate targets or biomarkers of the disease are nucleic acid sequences or proteins associated with the molecular feature which has been identified as having a causal relationship with the genetic risk variant and the disease feature. For example, where the molecular feature is the level of expression of a particular gene in the specific cell, the candidate target or biomarker can be the particular gene or the encoded protein, cis-acting nucleic acid sequences or trans-acting nucleic acid sequences or proteins.

[0067] The step of identifying candidate targets or biomarkers of the disease based on the one or more genetic risk variants identified as having a causal relationship with the one or more disease features is achieved by determining the location and / or association of the genetic risk variant. If the genetic risk variant is present within a particular gene, the candidate target / biomarker may be determined as the gene. For genetic risk variants that are located outside coding regions, the molQTL data, or other data, can be used to map the genetic variant to genes that are affected by the genetic risk variant.

[0068] In certain embodiments, hierarchical clustering is used to evaluate candidate targets or biomarkers. Distance methods for performing such hierarchical clustering include Euclidean, Pearson, Spearman, Manhattan, etc.

[0069] At step 110, a first cell type-specific knowledge graph is generated. The first cell typespecific knowledge graph comprises a first plurality of nodes, and each node represents one of the plurality of candidate targets or biomarkers, a biological pathway associated with the candidate targets or biomarkers, or one or more common interactors.

[0070] Here, a knowledge graph is a network, or graph, of the molecular pathways known to be associated with each of the candidate targets or biomarkers. The molecular pathways forming part of the knowledge graph include gene expression pathways and proteinprotein interaction pathways. A cell type-specific knowledge graph is specific for the cell type from which the molQTL data was obtained or to a group of cell types closely related to and encompassing the specific cell type or types from which the molQTL data was obtained, i.e., related cell subtypes having the ability to transform into the specific cell type from which the molQTL data was obtained. In certain embodiments, corresponding cell type-specific knowledge graphs are specific for the cell type from which the molQTL data was obtained.

[0071] A cell type-specific knowledge graph comprises nodes and edges connecting nodes for a specific cell type. A node may represent one of: one of the candidate targets or makers is (e.g., a gene identified as being causally related to the disease feature), a molecular pathway associated with the candidate targets or biomarkers, or a common interactor (e.g., genes which are related to a candidate target or biomarker and / or a molecular pathway associated with a candidate target or maker).

[0072] Cell type-specific knowledge graphs can be generated from open-source databases, such as ingenuity Pathways Analysis (IPA); Hetionet; CKG (Molecular quantitative trait loci. Nat Rev Methods Primers 3, 5 (2023); PrimeKG (Chandak, P., Huang, K. & Zitnik, M. Building a knowledge graph to enable precision medicine. Sci Data 10, 67 (2023)); etc. The cell type-specific knowledge graph can be created by tailoring the data to remove any genes or pathways that are not expressed in the cell type from which the molQTL data was obtained, namely genes that are not present with the cell type-specific transcriptomic profile. As will be described in more detail in relation to Figures 5A and 5B below, a cell type-specific knowledge graph associated with a specific cell type may be generated from a "parent" knowledge graph— i.e., a knowledge graph comprising genes and pathways that both are, and are not, expressed in the cell type from which the molQTL data was obtained. The parent knowledge graph is pruned (i.e., nodes and edges are removed) in accordance with a locality constraint such that the pruned parent knowledge graph is the cell typespecific knowledge graph. The parent knowledge graph is pruned in accordance with the locality constraint with respect to nodes within the parent knowledge graph that represent the candidate targets or biomarkers. As such, only those nodes within the parent knowledge graph which satisfy the locality constraint (in relation to the nodes representing the candidate targets or biomarkers) are maintained within the cell type-specific knowledge graph.

[0073] In one embodiment, the method 100 further comprises outputting (not shown) the cell type-specific knowledge graph.

[0074] Outputting the cell type-specific knowledge graph comprises storing the cell type-specific knowledge graph at a storage location of a computing device. The storage location may be a memory of a computing device such as a random access memory (RAM). Alternatively, the storage location may be a persistent storage location such as a non- transitory computer readable medium such as a hard disk drive, cloud storage location, etc. By storing the cell type-specific knowledge graph at a storage location of a computing device the predictive ability and the identification ability of the computing device is improved by virtue of the stored cell type-specific knowledge graph. Moreover, by pruning the cell type-specific knowledge graph from a "parent" knowledge graph (as described above), a compact representation of the pathways related to the candidate targets or biomarkers, or one or more common interactors, may be created and stored thereby making more efficient use of computing resources such as memory and storage requirements.

[0075] Additionally, or alternatively, outputting the cell type-specific knowledge graph comprises displaying a graphical representation of the first cell type-specific knowledge graph within a graphical user interface for review by a user. An example graphical representation is shown in Figure 7 which is described in more detail below. The graphical representation provides important visual feedback to a user regarding the relationships that have been identified within the cell type-specific knowledge graph that may not be identified or otherwise inferred from a non-graphical representation. Outputting such a graphical representation may therefore help improve the identification of key targets or biomarkers which may in turn help improve patient outcomes. The graphical representation is generated by applying a graph drawing algorithm to the cell type-specific knowledge graph such as a force direction drawing algorithm or a spring / repulsion model. The size of the nodes within the graphical representation may vary depending on interconnectivity (as described in relation to step 112 below) such that nodes having a high level of interconnectivity appear as larger within the graphical representation than nodes having a lower level of interconnectivity. Advantageously, this allows for improved identification of potential key targets or biomarkers by reducing the amount of display area given to, and thus reducing the visual influence of, nodes which are unlikely to be identified as being a key target or biomarker.

[0076] At step 112, a level of interconnectivity is calculated for each node within the first cell type-specific knowledge graph. A candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity is identified as a key target or biomarker.

[0077] The level of interconnectivity associated with a node is a measure of the relevance of the node within the overall context of the cell type-specific knowledge graph. A node having a high level of interconnectivity is indicative of the node (e.g., a gene, pathway, etc.) having a significant (e.g., greater) biological effect within the overall biology of the cell type, and especially with respect to a relevant biological effect, namely a biological effect predicted to have an effect in the disease feature. The level of interconnectivity, optionally with respect to a relevant biological effect, can therefore be used as a measure for ranking nodes and their relevance to a disease feature. The level of interconnectivity is determined for each node within the first cell type-specific knowledge graph. In one embodiment, the level of interconnectivity is a centrality measure for the node, such as a degree of the node or an eigencentrality of the node. The skilled person will appreciate that any suitable software package, such as NetworkX, may be used to calculate such measures. A candidate target or biomarker having a high level of interconnectivity, or a common interactor having a high level of interconnectivity, is identified as a key target or biomarker. As used throughout the present disclosure, a common interactor is a gene or protein present within the knowledge graph which has a high level of interconnectivity with the candidate targets or biomarkers but was not identified as a candidate target or biomarker. In certain embodiment, a common interactor is identified as a key target or biomarker.

[0078] The term "key target" refers to any target that can be targeted by a therapy in order to treat or prevent the disease. In certain embodiments, the key target is a gene sequence or protein. When the target is a gene sequence, gene therapy, including miRNA therapy, can be used to knockout or reduce expression of the gene sequence. When the target is a protein, a molecule capable of interacting with the protein to block or reduce its function, e.g., an antibody molecule or a small molecule, can be used. In some embodiments, the key target is not a known target for the disease. It may be a known target for a different disease, or it may not be known as a target for any disease. In certain embodiments the treatment can be provided in vivo or ex-vivo.

[0079] The term "a key biomarker" refers to any biomarker that can be used in the diagnosis of the disease and / or be used to monitor disease progression and / or severity. The key biomarker can be used on its own to diagnose the disease or can be used in combination with other biomarkers to diagnose the disease. The biomarker may be a gene sequence, a protein, or any other molecular trait. For example, the key biomarker can be the presence or level of a nucleic acid sequence or a protein; the level of methylation or phosphorylation of a nucleic acid; etc. In certain embodiments, the key biomarker is the presence or level of a nucleic acid sequence or a protein.

[0080] In certain embodiments, the target, biomarker, or common interactor having the highest degree of interconnectivity, and which is targetable or detectable, becomes the key target or biomarker. Other factors may also be considered in identifying a key target or biomarker, including whether the target or biomarker is unique to a specific cell type or conserved across cell types, whether the target or biomarker is causally associated with a plurality of disease features, the accessibility of the target or biomarker, and / or whether the target or biomarker is known or predicted to be associated with other diseases. For example, if a candidate target with high interconnectivity cannot be targeted therapeutically, the method of the present disclosure can be used to identify and rank druggable targets, i.e., candidate targets or common interactors having high interconnectivity while maintaining the desired cell type specificity. In one embodiment, the determination of whether the interconnectivity of a node is a high level of interconnectivity is based on the percentile levels of interconnectivity for all nodes within the cell type-specific knowledge graph. A node which has an interconnectivity level within an upper 75thpercentile of the levels of interconnectivity for all nodes within the cell type-specific knowledge graph is identified as a node representing a key candidate target or biomarker, or common interactor. In certain embodiments, a node which has an interconnectivity level within an upper 90thor 95thpercentile of the levels of interconnectivity for all nodes within the cell type-specific knowledge graph is identified as a node representing a key candidate target or biomarker, or common interactor. Additionally, or alternatively, the determination of whether the interconnectivity of a node is a high level of interconnectivity is based on a predetermined threshold. For example, if a node has a level of interconnectivity above a predetermined threshold value (e.g., 100, 150, 200 or more connections), then it is identified as a node representing a key candidate target or biomarker, or common interactor.

[0081] At optional step 114, a first candidate target or biomarker, or a first common interactor, is identified based on the level of interconnectivity of a respective node representing the first candidate target or biomarker, or the first common interactor.

[0082] At optional step 116, the first candidate target or biomarker, or the first common interactor, is output.

[0083] In one embodiment, outputting the first candidate target or biomarker, or the first common interactor, comprises storing, or saving, first candidate target or biomarker, or the first common interactor, to a persistent storage such as a non-volatile memory, a non-transitory medium, or the like. Additionally, or alternatively, outputting the first candidate target or biomarker, or the first common interactor, comprises transmitting the first candidate target or biomarker, or the first common interactor, via a network (e.g., a local area network, a wide area network, and the like), or displaying the first candidate target or biomarker, or the first common interactor, or a representation thereof, for review by a user.

[0084] The method 100 relates to an effective method for identifying a key target or a key biomarker for a disease. The key target can then be therapeutically targeted for treating the disease. The key biomarker can be used as a biomarker for disease diagnosis. The method 100 may additional comprise validating the key target or key biomarker by performing suitable in silico, in vitro or in vivo experiments. For example, in vitro or animal models of the disease can be treated by therapeutically targeting the key target. Samples or derived sample data from patients can be tested to confirm the effectiveness of a key biomarker in diagnosing or monitoring the disease.

[0085] Figure 2 shows steps which may be performed as part of the method 100, shown in Figures 1, according to an embodiment of the present disclosure.

[0086] Figure 2 shows the step 302 of obtaining a second cell type-specific knowledge graph, step 304 of generating a third cell type-specific knowledge graph, and step 306 of calculating levels of interconnectivity. Figure 2 further shows the optional step 310 of identifying a key target or maker, or common interactor, and optional step 312 of outputting the key target or biomarker, or common interactor. In one embodiment, the steps shown in Figure 2 are performed as part of the method 100 shown in Figure 1. That is, the steps of obtaining 302 and generating 304 may be performed after, or in conjunction with, the step of generating 110 in the method 100.

[0087] At step 302, a second knowledge graph comprising a second plurality of nodes is obtained. At least one node of the first plurality of nodes of the first cell type-specific knowledge graph (generated at the step of generating 110 in the method 100) and at least one node of the second plurality of nodes are identical or have a known biological relationship, e.g., are functionally related within a known biological pathway.

[0088] The second knowledge graph and the first cell type-specific knowledge graph comprise at least one node in common— that is, they both comprise at least one node representing a common molecular feature and / or pathway. The second knowledge graph is associated with cell type-specific molecular quantitative trait loci (molQTL) data for a second cell type. That is, the second knowledge graph is a cell type-specific knowledge graph associated with a cell type that is different to the cell type associated with the first cell type-specific knowledge graph. Alternatively, the second knowledge graph is associated with data for genetic risk variants associated with a second feature of the disease. That is, the second knowledge graph is a knowledge graph associated with a feature of the disease that is different to the first feature of the disease associated with the first cell type-specific knowledge graph.

[0089] In one embodiment, the second knowledge graph is obtained by generating the second knowledge graph using the process described above in relation to the method 100. Alternatively, the second knowledge graph is obtained by loading, receiving, or otherwise obtaining, the second knowledge graph from a computer readable medium.

[0090] At the step of generating 304, a third knowledge graph is generated. The third knowledge graph comprises the first cell type-specific knowledge graph and the second knowledge graph. The third knowledge graph comprises at least one edge connecting the at least one node of the first plurality of nodes and the at least one node of the second plurality of nodes.

[0091] The third knowledge graph is generated by "bridging" the first cell type-specific knowledge graph and the second knowledge graph to form a composition, or union, of the two graphs. To bridge the two graphs, a join can be formed by a common node present in the first and second knowledge graphs. Alternatively, the two graphs can be bridged by forming an edge between a node of the first cell type-specific knowledge graph and a node of the second knowledge graph wherein the nodes have a known biological relationship, e.g., are functionally related within a known biological pathway. The two graphs can also be bridged by using a common interactor between a node of the first cell type-specific knowledge graph and a node of the second knowledge graph.

[0092] When the second knowledge graph is associated with molQTL data for a second cell type, key targets or biomarkers can be identified from the third knowledge graph on a cell specific basis or across a number of cell types. This provides additional information and allows better identification of a desired key target or biomarker. For example, if a key target or biomarker (e.g., a gene or protein) is identified in multiple cell types, the present disclosure enables insights into which cell type should be prioritized, e.g., based on incidence and / or effect size of causal relationships between the key target or biomarker and the disease feature. Accordingly, the key targets or biomarkers may be identified on a cell specific basis or across a number of cell types.

[0093] When the second knowledge graph is associated with a second, different, feature of the disease, key targets or biomarkers can be identified within the third knowledge graph at a cell specific level which have a causal association with one or more disease features of a disease. This provides additional information and allows better identification of a desired key target or biomarker. For example, if a key target or biomarker (e.g., a gene or protein) is identified as having a causal association with a single or multiple disease features, the present disclosure enables insights into which target or biomarker should be prioritized, e.g., based on which disease feature or features are to be targeted, incidence and / or effect size of causal relationships between the key target or biomarker and each of the disease features. Accordingly, key targets or biomarkers may be identified with respect to one or more disease features and / or on a cell specific basis or across a number of cell types. When the second knowledge graph is associated with a second, different, feature of the disease, it is inherent that at least one of the features of the disease is not disease incidence, i.e., presence of the disease. In certain embodiments, bot the first and the second feature of the disease are not disease incidence, i.e., presence of the disease.

[0094] 11 The step 306 of calculating levels of interconnectivity corresponds to step 112 for calculating levels of interconnectivity in the method 100, but with interconnectivity being calculated for each node within the third knowledge graph. As such, the reader is referred to the description of step 112 for further details as to the operations performed at the calculating step 306. Similarly, the optional steps of identifying 310 and outputting 312 correspond to the steps to identifying 114 and outputting 116 as described in relation to the method 100 (but using the third knowledge graph).

[0095] The skilled person will appreciate that the above-described steps of Figure 2 may be repeated for two or more cell types and / or two or more disease features such that the key targets or biomarkers are identified from a knowledge graph comprising compositions of several cell type-specific knowledge graphs and / or knowledge graphs associated with several features of disease. In one embodiment, cell type-specific disease feature-specific knowledge graphs are generated for each combination of cell type and disease feature. For example, where the molQTL data is obtained for two different cell types (e.g., cell types A and B), and genetic risk variant data is obtained for two different disease features (e.g., diseases feature X and Y), cell type-specific disease feature-specific knowledge graphs are generated (using the approach described above in relation to the method 100 shown in Figure 1) for: cell type A and disease feature X; cell type A and disease feature Y; cell type B and disease feature X; and cell type B and disease feature Y. These knowledge graphs may be analyzed separately to identify key targets or biomarkers for each cell type and disease feature combination or combined using the approach described in relation to Figure 2. The combined, or bridged, knowledge graph may be used to identify a key candidate target or biomarker, or common interactor.

[0096] Figure 3 illustrates 2-sample mendelian randomization (2SMR).

[0097] Figure 3 shows an instrument variable 402, Z, an exposure 404, X, an outcome 406, Y, and one or more confounders 408, U. The instrument variable 402, Z, corresponds to a genetic risk variant (e.g., a single single-nucleotide polymorphism / variation, SNP / SNV) associated with a feature of a disease. The exposure 404, X, corresponds to one or more molecular features and the outcome 406, Y, corresponds to the disease feature. The association between the genetic risk variant, Z, and the molecular features, X, is represented as (ZX. The association between the genetic risk variant, Z, and the disease feature, Y, is represented as pZY. The association between the molecular features, X, and the disease feature, Y, is represented as pXY.

[0098] The associations pzxare estimated from a first dataset (e.g., the data obtained at the obtaining step 102 of the method 100) whilst the associations pZYare estimated from a second dataset (e.g., the data obtained at step 104 of the method 100). As the two datasets comprise summary data, 2SMR is used to estimate the causal relationship between the molecular features, X, and the disease feature, Y— i.e., estimate the association pXY.

[0099] Thus, 2SMR is used to test the causal hypothesis that the molecular features, X, are causally related to the disease feature, Y, whilst also providing an estimate for the causal effect (i.e., the size or degree of the positive or negative causal effect that the molecular feature has on the disease feature).

[0100] As stated above, 2SMR may be used as part of the identifying step 106 in the method 100 or the identifying step 206 in the method 200. Any suitable software package or library may be used to perform 2SMR such as the TwoSampleMR module of the R programming language.

[0101] Figure 4 shows a matrix 502 of causal relationships according to embodiments of the present disclosure.

[0102] The matrix 502 (alternatively referred to as a causal matrix, effect matrix, or causal relationship / effect matrix) comprises a plurality of causal effect values determined in relation to a plurality of cell types (horizontal axis of the matrix 502) and a plurality of molecular features (vertical axis of the matrix 502) and a disease feature or outcome (e.g., motor score severity for Parkinson's disease). Alternatively, but not shown in Figure 4, the matrix may comprise a plurality of causal effect values determined in relation to a plurality of disease features or outcomes (e.g., the horizontal axis of the matrix) and a plurality of molecular features (e.g., the vertical axis of the matrix) and a specific cell type (e.g., a specific T cell subtype). The magnitude of the causal relationship is represented by the darkness of the shading of the various causal effect values - no shading indicates no causal relationship, whereas dark shading indicates a strong causal relationship.

[0103] Figure 4 shows a first value 504 of the matrix 502, a second value 506 of the matrix 502, a third value 508 of the matrix 502, and a fourth value 510 of the matrix 502.

[0104] In Figure 4, each value within the matrix 502 may be understood as quantifying the causal relationship that a cell type-molecular feature combination has on the disease outcome. The first value 504 corresponds to a measure of the causal relationship between the cell type, CTi, and the molecular feature, MFi, and the disease feature or outcome. For example, the first value 504 may correspond to the size of the causal relationship between the gene CCDC92(MFi) of the Thl / 17 cell (CTi) and the Parkinson's disease motor score severity (UPDRS III). The second value 506 corresponds to a measure of the causal relationship between the cell type, CTi, and the molecular feature, MF / , and the disease feature or outcome. The third value 508 corresponds to a measure of the causal relationship between the cell type, CT, and the molecular feature, MF / , and the disease feature or outcome. The fourth value 510 corresponds to a measure of the causal relationship between the cell type, CTn, and the molecular feature, MF / , and the disease feature or outcome.

[0105] The first value 504 and the fourth value 510 are approximately equal to 0, meaning that MFi of CTi and MF / of CTnare not causally related to the disease feature or outcome. The second value 506 is negative meaning that MF / of CTi is causally related and negatively associated with the risk of developing / having the disease feature or outcome. The magnitude of the second value 506 represents the magnitude of the causal relationship between MF / , CTi and the disease feature or outcome. The third value 508 is positive meaning that MF / of CT / is causally related to and positively associated with the risk of developing / having the disease feature or outcome. The magnitude of the third value 508 represents the magnitude of the causal relationship between MF / , CT / and the disease feature or outcome.

[0106] The matrix 502 may be generated by iteratively applying a causal analysis approach (e.g., mendelian randomization) for all cell type - molecular feature pairs in relation to a given disease feature or outcome. As stated above in relation to Figure 1, a matrix such as the matrix 502 may be used to identify candidate targets or biomarkers. For example, the molecular feature, MF / , may be identified as a candidate target or biomarker in relation to the cell type, CT / , and the disease feature or outcome because the third value 508 is high in relation to other values within the column of values associated with CT / . Here, a high causal effect value for a cell type is determined by identifying the largest causal effect value across all molecular features for the cell type. Alternatively, a high causal effect value is determined by identifying any causal effect values above a predetermined threshold across all module features for the cell type.

[0107] Figure 5A shows a knowledge graph 600 according to an embodiment of the present disclosure.

[0108] The knowledge graph 600 comprises a plurality of nodes including nodes 602-614 representing molecular features (e.g., genes), nodes 616-622 representing molecular pathways, node 624 representing a first candidate target or biomarker, and node 626 representing a second candidate target or biomarker. Node 624 and node 626 are both / representing molecular features which have been identified as candidate targets or biomarkers.

[0109] 9 The knowledge graph 600 corresponds to a "parent" knowledge graph from which a cell type-specific knowledge graph is generated. As such, the knowledge graph 600 comprises genes and pathways that are, and are not, expressed in the cell type to which the cell type-specific knowledge graph relates. To generate the cell type-specific knowledge graph, the knowledge graph 600 is pruned to remove any nodes which are not biologically associated / annotated to the cell type and candidate targets or biomarkers. The pruning step is used to remove genes, proteins and pathways from the parent knowledge graph that are not annotated in the specific cell type. The pruning step could also be used to remove genes, protein and pathways that are active in the specific cell type but are not of interest because they are active in different unrelated cell types and / or are involved in cellular homeostasis. The knowledge graph 600 is also pruned in accordance with one or more locality constraints. For example, nodes within the knowledge graph 600, which are identified as not representing the candidate targets or biomarkers, or are not within a certain distance of the candidate targets or biomarkers, are removed by pruning. A cell type-specific knowledge graph is generated from the knowledge graph 600 by pruning, or removing, any nodes which are not identified and / or do not satisfy the one or more locality constraints. Advantageously, the locality constraints help to ensure that there is a relationship between the nodes that is not significantly affected by numerous intermediate or bystander nodes.

[0110] A first locality constraint is based on a shortest path constraint such that a shortest path distance between pairs of nodes representing candidate targets or biomarkers within the cell type-specific knowledge graph is less than a predetermined distance (e.g., less than 4 or less than 3). In the example shown in Figure 5A, nodes within the knowledge graph 600 which represent candidate targets or biomarkers are identified (i.e., node 624 and node 626) and the shortest paths between each pairing of these nodes are calculated. There are two shortest paths between node 624 and node 626, i.e., the path involving node 616 and the path involving node 620. Because the shortest paths between node 624 and node 626 are less than 4 (that is, there are fewerthan 4 nodes on the paths connecting node 624 and node 626) the nodes on these paths are maintained within the cell typespecific knowledge graph.

[0111] A second locality constraint is based on the direct neighbors of nodes representing candidate targets or biomarkers. That is, any node within the knowledge graph 600 which is a direct neighbor (i.e., is directly connected to) a node representing a candidate target or biomarker is maintained within the cell type-specific knowledge graph. In the example shown in Figure 5A, the application of the second locality constraint (in addition to the first locality constraint) would result in nodes 602, 604, 614, 618 being maintained within the

[0112] 15 cell type-specific knowledge graph because they are all direct neighbors of node 624 or node 626.

[0113] A third locality constraint assesses the specificity and significance of nodes in the knowledge graph using 1-sample permutation analysis. More particularly, 1-sample permutation analysis is performed based on node degree and the corresponding false discovery rate (FDR) computed by comparing the node degree in cell type-specific knowledge graph against a random graph. Any nodes having a p-value < 0.05 are maintained within the cell type-specific knowledge graph (i.e., any node having a p-value > 0.05 is pruned from the cell type-specific knowledge graph).

[0114] Nodes within the knowledge graph 600 which do not satisfy the above one or more locality constraints are pruned, or removed, resulting in a cell type-specific knowledge graph being generated.

[0115] Figure 5B shows a cell type-specific knowledge graph 600-1 generated from the knowledge graph 600 shown in Figure 5A according to an embodiment of the present disclosure.

[0116] The cell type-specific knowledge graph 600-1 corresponds to the knowledge graph 600 shown in Figure 5A pruned in accordance with the one or more locality constraints such that the nodes within the knowledge graph 600 which do not satisfy the one or more locality constraints are removed. That is, nodes 606-614 and node 622 within the knowledge graph 600 have been pruned and do not appear within the cell type-specific knowledge graph 600-1. As such, the cell type-specific knowledge graph 600-1 comprises nodes having a close, biologically relevant and statistically significant connection with the candidate targets or biomarkers.

[0117] A level of interconnectivity may then be calculated for each node within cell type-specific knowledge graph 600-1 to identify possible key targets or biomarkers. As stated above, the degree of a node may be used as a measure for the level of interconnectivity of the node. In the cell type-specific knowledge graph 600-1 nodes 602, 614, and 618 have degree of 1, nodes 604, 616, 620, and 626 have degree of 3, and node 624 has degree of 5. Node 624 may thus be identified as a key target or biomarker because it has a high level of interconnectivity (relative to the level of interconnectivity of other nodes within the cell type-specific knowledge graph 600-1).

[0118] In certain embodiment, genetic risk variant data can additionally be obtained for one or more comorbidities. The genetic risk variant data for the one or more comorbidities is determined to have a causal relationship with the one or more comorbidities and one or more associated molecular features in the same manner as described above for the one or more disease features. This step enables the identification of a plurality of candidate targets or biomarkers of the one or more comorbidities. A knowledge graph of the molecular pathways known to be associated with each of the candidate targets or makers is generated in the manner described above. The knowledge graph for the comorbidity can then be combined with one or more knowledge graphs obtained for features of the disease in the manner described above. Accordingly, key targets or makers can then be identified based on one or more cell types, one or more disease features, and on a comorbidity. Such a method provides further information on key targets or biomarkers for a disease that has an associated comorbidity.

[0119] A comorbidity is the presence of an additional condition that presents itself along with the primary disease. Examples include depression in combination with a diabetic comorbidity, or Parkinson's disease in combination with a diabetic comorbidity.

[0120] EXAMPLE IMPLEMENTATION

[0121] Data sources

[0122] All data used in the 2SMR analyses were from published articles: Schmiedel et al., Cell, 175, 1701-1715, 2018; Nalls et al., Lancet Neurol., 18, 1091-1102, 2019; Blauwendraat et al., Mov Disord., 34, 866-875, 2019; Iwaki et al., Mov Disord., 34, 1839-1850, 2019; and Liu et al., Nature Genetics, 47, 979-986, 2015. Additionally, data used to build the KG were from public databases (i.e., Entrez Gene, Gene Ontology, Wikipathways, Reactome, and / or Disease Ontology) and extracted relationships from Hetionet and / or PrimeKG.

[0123] Immune cell type-specific eOTL summary statistics

[0124] Filtered eQTL data for 15 purified human immune cell types (including B, CD4 naive, CD4 stimulated, CD8 naive, CD8 stimulated, monocytes classical, monocytes nonclassical, NK, TFH, TH1, TH2, TH1 / 17, TH17, Treg naive, and Treg memory cells) were available in variant call format (VCF) through the DICE database. All data were mapped to GRCh37 human reference genome. Listed genetic polymorphisms with significant associations to the expression of a nearby gene(s) were previously filtered for significance as follows: adjusted p-value < 0.05, raw p-value < 0.0001, and transcript per million (TPM) > 1.0. For more information on how these data were generated, see Schmeidel et al., 2018 (supra).

[0125] In preparation to use these immune cell type-specific eQTL as exposure variables in 2SMR analyses, the following data fields were extracted from all VCFs using python (version

[0126] 9 3.8.8): rsID, ensemblelD, geneSymbol, pvalue (p < 0.0001), beta, stat, and FDR. Additionally, since the beta and stat (t statistics) were provided, the standard error (SE) was calculated as follows: me$all$eqtls$beta_se ~ me$all$eqtls$beta / me$all$eqtls$statistic

[0127] Extracted eQTL summary statistics were saved in csv format.

[0128] GWAS summary statistics

[0129] All GWAS summary statistics were processed using R (versions 3.6.1 - 4.2.2) through IDE RStudio (v2021.09.0 - v2022.07.2) unless otherwise specified. Although manual curation was required for two out of three datasets (described below), all final GWAS summary statistic outputs were mapped to GRCh37 human reference genome and were formatted following IEU OpenGWAS Project guidelines, which include the following 16 data fields: rsID, chromosome, position, beta, SE, sample size, pvalue, effect allele frequency (EAF), effect allele, other allele, outcome, outcome ID, original name outcome, deprecated outcome, MR keep outcome, and data source. All final GWAS summary statistics were saved in csv format. The following describes processing used to generate final GWAS summary statistics outputs.

[0130] Parkinson's disease (PD) incidence (Nalls et al., 2019 (supra)) and inflammatory bowel disease incidence (Liu et al., 2015(supra)) GWAS summary statistics were processed directly through IEU OpenGWAS Project using R modules TwoSampleMR (version 0.5.6) and ieugwasr (version 0.1.5). Specifically, statistically significant SNPs (p < 0.0001) and their associated data fields were extracted from PD dataset "ieu-b-7" using an ieugwasr access token.

[0131] PD age at onset GWAS summary statistics (Blauwendraat et al., 2019 (supra)) were available for local download as a txt file through the International Parkinson's disease Genomics Consortium. Of the 15 original data fields provided, the following seven fields were selected for further processing: MarkerName (concatenated chromosome and position information for each SNP), Allele 1 (effect allele confirmed by authors), Allele 2, Freql (EAF), Effect (beta), StdErr (SE), and P-value. Only entries with P-value < 0.0001 were kept. Allele 1 and 2 data entries were reformatted to uppercase lettering. Chromosome and position fields were extracted from "MarkerName." Sample size, outcome, outcome ID, original name outcome, deprecated outcome, MR keep outcome, and data source fields were added manually to reflect information provided in the original publication. Since the original GWAS summary statistics data file did not include rsIDs or reference alleles (i.e., other alleles) information, chromosome and position coordinates were lifted over from GRCh37 to GRCh38 human reference genome using CrossMap python function (version 3.8.8), so that rsID and reference allele identification was compatible with the following four methods. In the first method, using dbSNP GRCh38 reference file as a dictionary, rsIDs, reference alleles, and their associated GRCh38 coordinates were identified for the majority of SNPs using python function. In the second method, where rsIDs were not found in the dbSNP GRCh38 reference file, rsIDs were identified using R module SNPIocs.Hsapiens.dbSNP151.GRCh38 (version 0.99.23). In the third method, where reference alleles were not found in the dbSNP GRCh38 reference file, reference alleles were identified using R module RMySQL (version 0.10.25). Finally, for three SNPs whose rsIDs were not found by either method, information was entered manually by crossreferencing dbSNP and Ensembl VEP websites.

[0132] PD UPDRSIII GWAS summary statistics (Iwaki et al., 2019 (supra)) were available for local download as a txt file through the PD Progression Meta-data GWAS Browser. Of the seven original data fields provided, the following five fields were selected for further processing: SNP (concatenation of chromosome and position coordinates), BETA, SE, P, and N (sample size). Only entries with P < 0.0001 were kept. Chromosome and position fields were extracted from "SNP." Sample size, outcome, outcome ID, original name outcome, deprecated outcome, MR keep outcome, and data source fields were added manually to reflect information provided in the original publication. Like Blauwendraat et al., 2019 (supra), original GWAS summary statistics, rsIDs and reference alleles (as well as effect alleles) were not provided in Iwaki et al., 2019 (supra). Therefore, these data were identified after lifting over genomic coordinates from GRCh37 to GRCh38 human reference genome using one or more of the following methods: (1) cross-reference dbSNP GRCh38 reference file, (2) use R module SNPIocs.Hsapiens.dbSNP151.GRCh38, or (3) manual curation using dbSNP and Ensembl VEP websites.

[0133] Two sample mendelian randomization (2SMR)

[0134] 2SMR analyses were completed on each of 15 immune cell types across 3 disease features iteratively using R module TwoSampleMR. Briefly, for each immune cell type by disease feature combination, any shared SNPs were identified and stored as a respective list. Next, immune cell extracted eQTL csv files were filtered based on shared SNP lists and saved as a new csv file (e.g., PD_incidence_x_B_cell_naive_extracted_se.csv). Lastly, newly generated clinical_outcome_x_immune_cell_extracted_se.csv files were read in as exposure data, data were harmonized with disease feature GWAS summary statistics, and 2SMR analyses were run. Immune cell type-specific labels were added to each result summary, which included the following 11 data fields: exposure (gene symbol), id. exposure (gene symbol), id. outcome, outcome, method, nsnp, b (beta), se, pval, cell type, and SNP (rsID). Given only single SNPs were identified as viable instruments per gene of interest across all runs, the wald ratio method was used. Finally, for a disease feature of interest, immune cell type-specific results were merged and saved as a csv file.

[0135] 2SMR visualization

[0136] To visualize 2SMR results per clinical outcome as a heatmap, merged immune cell typespecific results (including variables exposure, cell type, and b) were reshaped into a wide matrix. Any "NA" values were replaced with a "0" as a holder so that hierarchical clustering could be applied. Heatmaps were generated using the R module ComplexHeatmap (version 2.14.0).

[0137] Figures 6A-6D show the causal effect matrices— i.e., 2SMR visualizations— obtained for the PD disease features / outcomes of incidence (Figure 6A), age of onset (Figure 6B), and motor score severity (Figure 6C), as well as inflammatory bowel disease incidence (Figure 6D).

[0138] As indicated above, the use of cell type-specific molecular quantitative trait loci (molQTL) data enables accurate causal effects to be identified at a cell type-specific level. For example, if molecular quantitative trait loci (molQTL) data from whole blood is used (instead of using cell type-specific data), for a target that is causally associated with multiple cell types, an average of its effect size will be seen, which will be driven by the most robust effect, and can hide differences in magnitude and / or directionality of that association. For example, in PD incidence 2SMR and KANSL1 had positive and negative causal associations, respectively, with respect to the risk of developing PD depending on cell type. The relevant data extracted from Figure 6A in respect of gene KANLS1 for 6 different cell types, as well as the overall average, is provided below in Table II.

[0139] TABLE II

[0140] T cell lineage-specific knowledge graphs

[0141] The largest number and most robust causal relationships between T cells and the three PD disease features were identified - therefore T cells (including CD4 naive, CD4 stimulated, CD8 naive, CD8 stimulated, TFH, TH1, TH2, TH1_17, TH17, Treg naive, and Treg memory cells) were prioritized for further evaluation by knowledge graph (KG) analysis. Although individual T cell type KGs can be generated and used in the analysis, it was decided that l evaluating all T cells would provide more insights, as many T cell effector cells have the capability / plasticity to change between subtypes depending on the microenvironment (as compared to other immune cell type lineages). To create a T cell lineage-specific KG, gene features from open source databases (e.g., Entrez Gene, Gene Ontology, Reactome, etc.) were removed if said gene was not present in any T cell type-specific transcriptomic profile (see that disclosed in Schmeidel et al., 2018 (supra)) using Hetionet or PrimeKG as a KG framework but focusing only on the 'Biological Process', 'Cellular Component', 'Disease', 'Gene', 'Molecular Function' and / or 'Pathway' meta-nodes and their corresponding edges. For example, using this T cell lineage-specific knowledge graph, we then created a PD motor score severity (UPDRSIII scoring) disease feature KG and inputted the 21 causal genes identified from the 2SMR analysis; however, only 12 of these genes were identified within our KG (9 genes were not found in the KG used in this example).

[0142] Cell type-specific knowledge graph analytics and visualization

[0143] Graph-related calculations were generated using NetworkX (version 2.7) from Python (version 3.10). To generate the cell specific disease feature network of connectivity for the one or more disease features, the shortest path between pairs of 2SMR result input genes (i.e., candidate targets / biomarkers) was calculated and the analysis restricted to only include first neighbors or nodes connecting 2 causal genes with a path length < 3. The resulting network shows the connectivity between the genes (nodes). To assess the specificity and significance of genes (nodes) in this graph, 1-sample permutation analysis was performed based on node degree and the corresponding false discovery rate (FDR) computed by comparing the node degree in final graph against a random graph. Finally, the network was further refined by containing only nodes with p-value < 0.05, and node degree was computed to determine which nodes were most interconnective within the graph. Graphs were visualized using Gephi (version 00.9.7 202208031831).

[0144] Figure 7 shows a visualization of the cell type-specific knowledge graph using the abovedescribed techniques applied to a single disease feature. Figure 8 shows a visualization of the cell type-specific knowledge graph using the above-described techniques applied to multiple disease features.

[0145] Results

[0146] The method provides a cell type-specific, directionality informed process for identifying candidate targets / biomarkers with causal links to a disease feature (e.g., PD patient clinical features). In total, 126 unique candidate targets / biomarkers across three PD disease features - incidence, age of onset, and motor score severity - were identified, including resolving the directionality of the target / biomarker (positive or negative effect size) and the association to one or more specific immune cell populations (Figure 6A-C). Additionally, 264 unique immune cell-specific and directionality informed candidate targets / biomarkers for inflammatory bowel disease (IBD) were identified - of which 7 candidate targets were shared with one or more PD disease feature 2SMR results (Figure 6D).

[0147] In the method, the causal genes (i.e., candidate targets / biomarkers) were contextualized within a cell type-specific KG to identify common interactors (i.e., additional candidate targets / biomarkers), identify enriched pathways, and rank the targets / biomarkers based on their degree of interconnectivity. For example, in T cell type-specific KG for PD motor score severity, 12 causal genes were inputted and determined to form a network comprising 80 KG-genes as well as 49 biological pathways (see Figure 7). Top ranking genes had 30+ connections to other nodes within the KG and are determined to be a target or biomarker of PD. In an additional example, when evaluating shared biology between IBD and PD disease features, 13 genes (1 derived from 2SMR runs and 12 additional found through KG) and 5 biological pathways were identified (see Figure 8).

[0148] The method contextualizes the 2SMR results to derive additional biological insights within a cell type-specific KG and thereby identifies broader biological changes in pathways and / or molecular functions (rather than focusing on single gene / protein candidate targets as the primary output). This enables the identification of targets / biomarkers that have a high degree of interconnectivity demonstrating with high confidence the involvement of the targets / biomarkers with the disease. Such targets / biomarkers can therefore be considered to be strong, i.e., key, targets / biomarkers of the disease. In addition, due to the identification of the broader biological changes in pathways and / or molecular functions, common interactors can be identified that have a high degree of interconnectivity to the 2SMR results and can also be considered with high confidence to be targets / biomarkers of the disease.

[0149] These relationships are graphically depicted by Figure 8, which illustrates an example of graphical output for display by a graphical user interface (GUI) or a display screen. The output (as demonstrated by Figure 8) further represents a compact and informative representation of a cellular network related to a candidate target or marker, which may be stored efficiently in a computer memory (e.g., memory 906), as a cell type-specific knowledge graph. In some aspects, the cell type-specific knowledge graph may comprise linked list, one or more database tables linked by key values, or other storage links designed to reduce memory utilization of the computer memory for efficient storage thereon. As shown in the example of Figure 8, the related input for the cell type-specific knowledge graph was PD incidence, PD age of onset, PD motor score severity, and IBD. As shown for Figure 8, each of these values are linked via one or more genes or pathways to form the compact and informative representation of the cellular network.

[0150] Figure 9 shows an example computing system for carrying out the methods of the present disclosure. Specifically, Figure 9 shows a block diagram of an embodiment of a computing system according to example embodiments of the present disclosure. The computing system shown in Figure 9 may correspond to a part, or the whole, of any of the functional units described above.

[0151] Computing system 900 can be configured to perform any of the operations disclosed herein such as, for example, any of the operations discussed with reference to the methods of Figures 1 and 2. Computing system includes one or more computing device(s) 902. The one or more computing device(s) 902 of computing system 900 comprise one or more processors 904 and memory 906. One or more processors 904 can be any general purpose processor(s) configured to execute a set of instructions. For example, one or more processors 904 can be one or more general-purpose processors, one or more field programmable gate array (FPGA), and / or one or more application specific integrated circuits (ASIC). In one embodiment, one or more processors 904 include one processor. Alternatively, one or more processors 904 include a plurality of processors that are operatively connected. One or more processors 904 are communicatively coupled to memory 906 via address bus 908, control bus 910, and data bus 912. Memory 906 can be a random access memory (RAM), a read only memory (ROM), a persistent storage device such as a hard drive, an erasable programmable read only memory (EPROM), and / or the like. The one or more computing device(s) 902 further comprise I / O interface 914 communicatively coupled to address bus 908, control bus 910, and data bus 912.

[0152] Memory 906 can store information that can be accessed by one or more processors 904. For instance, memory 906 (e.g., one or more non-transitory computer-readable storage mediums, memory devices) can include computer-readable instructions (not shown) that can be executed by one or more processors 904. The computer-readable instructions can be software written in any suitable programming language or can be implemented in hardware. Additionally, or alternatively, the computer-readable instructions can be executed in logically and / or virtually separate threads on one or more processors 904. For example, memory 906 can store instructions (not shown) that when executed by one or more processors 904 cause one or more processors 904 to perform operations such as any of the operations and functions for which computing system 900 is configured, as described herein. In addition, or alternatively, memory 906 can store data (not shown) that can be obtained, received, accessed, written, manipulated, created, and / or stored. In some implementations, the one or more computing device(s) 902 can obtain from and / or store data in one or more memory device(s) that are remote from the computing system 900.

[0153] Computing system 900 further comprises storage unit 916, network interface 918, input controller 920, and output controller 922. Storage unit 916, network interface 918, input controller 920, and output controller 922 are communicatively coupled to the central control unit (i.e., the memory 906, the address bus 908, the control bus 910, and the data bus 912) via I / O interface 914.

[0154] Storage unit 916 is a computer readable medium, preferably a non-transitory computer readable medium, comprising one or more programs, the one or more programs comprising instructions which when executed by the one or more processors 904 cause computing system 900 to perform the method steps of the present disclosure. Alternatively, storage unit 916 is a transitory computer readable medium. Storage unit 916 can be a persistent storage device such as a hard drive, a cloud storage device, or any other appropriate storage device.

[0155] Network interface 918 can be a Wi-Fi module, a network interface card, a Bluetooth module, and / or any other suitable wired or wireless communication device. In an embodiment, network interface 918 is configured to connect to a network such as a local area network (LAN), or a wide area network (WAN), the Internet, or an intranet.

[0156] The above illustrative examples of various aspects and implementations provide an overview for understanding aspects and implementation of the disclosed method. The figures provided herein depict exemplary aspects of the present system and methods and are not intended to limit the scope of the disclosure.

[0157] Unless otherwise stated, all technical terms used herein have the same meaning as commonly understand by a person skilled in the art. Singular forms "a", "an" and "the" include plural references unless the context of the disclosure clearly dictates otherwise. The term "or" is intended to encompass "and / or" unless clearly stated otherwise.

[0158] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. i ASPECTS OF THE PRESENT DISCLOSURE

[0159] The following aspects of the disclosure are exemplary only and not intended to limit the scope of the disclosure.

[0160] Aspect 1. A method of identifying key targets or biomarkers for a disease, the method comprising : obtaining, by one or more processors, data for genetic risk variants associated with a first feature of the disease; obtaining, by the one or more processors, cell typespecific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature; determining, by the one or more processors, one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features; identifying, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generating, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculating, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell typespecific knowledge graph is identified as a key target or biomarker.

[0161] Aspect 2. The method of aspect 1 further comprising: outputting, by the one or more processors, the first cell type-specific knowledge graph.

[0162] Aspect 3. The method of aspect 2 wherein the step of outputting the first cell type-specific knowledge graph comprises storing the first cell type-specific knowledge graph at a storage location of a computing device.

[0163] Aspect 4. The method of aspect 2 wherein the step of outputting the first cell type-specific knowledge graph comprises displaying a graphical representation of the first cell typespecific knowledge graph within a graphical user interface.

[0164] Aspect 5. The method of any one of the proceeding aspects further comprising: identifying, by the one or more processors, a first candidate target or biomarker, or a first common interactor, based on the level of interconnectivity of a respective node representing the first candidate target or biomarker, or the first common interactor.

[0165] 15 Aspect 6. The method of aspect 5 further comprising: outputting, by the one or more processors, the first candidate target or biomarker, or the first common interactor.

[0166] Aspect 7. The method of any one of the proceeding aspects wherein the causal relationship is determined from the data for genetic risk variants and the molQTL data using mendelian randomization.

[0167] Aspect 8. The method of any one of the proceeding aspects, further comprising: obtaining, by the one or more processors, a second knowledge graph comprising a second plurality of nodes, wherein at least one node of the first plurality of nodes and at least one node of the second plurality of nodes are related.

[0168] Aspect 9. The method of aspect 8 wherein the second knowledge graph is associated with cell type-specific molecular quantitative trait loci (molQTL) data for a second cell type.

[0169] Aspect 10. The method of aspect 8 wherein the second knowledge graph is associated with data for genetic risk variants associated with a second feature of the disease.

[0170] Aspect 11. The method of aspect 8 further comprising: generating, by the one or more processors, a third knowledge graph comprising the first cell type-specific knowledge graph and the second knowledge graph, wherein the third knowledge graph comprises at least one edge connecting the at least one node of the first plurality of nodes and the at least one node of the second plurality of nodes, wherein a candidate target or biomarker, or a common interactor, represented by a node within the third knowledge graph having a high level of interconnectivity is identified as a key target or biomarker.

[0171] Aspect 12. The method of aspect 11 wherein each at least one edge connects nodes representing a related candidate target or biomarker, related molecular pathway, or related common interactor.

[0172] Aspect 13. The method of any one of the proceeding aspects wherein the first cell typespecific knowledge graph is generated from a parent knowledge graph.

[0173] Aspect 14. The method of aspect 13 wherein the first cell type-specific knowledge graph is generated by pruning the parent knowledge graph in accordance with a locality constraint with respect to nodes within the parent knowledge graph representing the candidate targets or biomarkers.

[0174] Aspect 15. The method of aspect 14 wherein the locality constraint is a shortest path constraint such that a shortest path distance between pairs of nodes representing candidate targets or biomarkers within the first cell type-specific knowledge graph is less than a predetermined distance.

[0175] Aspect 16. The method of aspect 15 wherein the predetermined distance is less than 4.

[0176] Aspect 17. The method of any one of the proceeding aspects wherein the level of interconnectivity of a node is one of: a degree of the node, or an eigencentrality of the node.

[0177] Aspect 18. The method of any one of the proceeding aspects wherein a node is determined to have a high level of interconnectivity if the level of interconnectivity for the node is within an upper quartile of levels of interconnectivity for all nodes within the first cell typespecific knowledge graph.

[0178] Aspect 19. The method of any one of the proceeding aspects wherein nodes within the first cell type-specific knowledge graph associated with molecular pathways include nodes representing gene expression pathways and / or nodes representing protein-protein interaction pathways.

[0179] Aspect 20. The method of any one of the proceeding aspects wherein the key target is a target that can be targeted by a therapy in order to treat or prevent the disease.

[0180] Aspect 21. The method of aspect 20 wherein the key target is a nucleic acid sequence or a protein.

[0181] Aspect 22. The method of any one of the proceeding aspects wherein the key biomarker for a disease is indicative of the presence or severity of the disease.

[0182] Aspect 23. The method of aspect 22 wherein the key biomarker is presence or level of a nucleic acid sequence or a protein, or the level of methylation or phosphorylation of a nucleic acid.

[0183] Aspect 24. The method of any one of the proceeding aspects wherein the genetic risk variants associated with the first feature of the disease are: single-nucleotide polymorphisms / variations (SNPs / SNVs); insertion and / or deletion variations; copy number variations; translocation and / or an inversions, or a combination thereof.

[0184] Aspect 25. The method of aspect 24 wherein the genetic risk variants are SNPs.

[0185] Aspect 26. The method of any one of the proceeding aspects wherein the disease is: heart disease, such as ischemic heart disease and coronary artery disease; cancer; a bacterial or viral infection; a cerebrovascular disease, such as stroke; a respiratory disease, such l as chronic obstructive pulmonary disease; diabetes; an autoimmune disease, such as ulcerative colitis, Crohn's disease, inflammatory bowel disease, rheumatoid arthritis, Guillain-Barre syndrome, Sjogren's syndrome, scleroderma and Graves' disease; or a neurological disorder, such as Alzheimer's disease, multiple sclerosis, Parkinson's disease, amyotrophic lateral sclerosis (ALS) and frontotemporal dementia.

[0186] Aspect 27. The method of any one of the proceeding aspects wherein the molQTL data associates a genetic variant with a molecular feature and the genetic variant is selected from the group consisting of: single-nucleotide polymorphisms / variations (SNPs / SNVs); insertion and / or deletion variations; copy number variations; translocation and / or an inversions, or a combination thereof.

[0187] Aspect 28. The method of aspect 27 wherein the genetic risk variant is a SNP.

[0188] Aspect 29. The method of any one of the proceeding aspects wherein the molQTL data is cell type-specific expression quantitative trait loci (eQTL) data, cell type-specific DNA methylation quantitative trait loci (meQTL) data, cell type-specific histone modification quantitative trait loci (hQTL) data, cell type-specific chromatin accessibility quantitative trait loci (caQTL) data, cell type-specific alternative gene splicing quantitative trait loci (sQTL) data, cell type-specific protein level quantitative trait loci (pQTL) data, cell typespecific microRNA expression quantitative trait loci (mirQTL) data or cell type-specific ribosome occupancy quantitative trait loci (rQTL) data.

[0189] Aspect 30. The method of any one of the proceeding aspects wherein the data for the one or more genetic risk variants is harmonized with the molQTL data prior to the step of determining the causal relationship.

[0190] Aspect 31. The method of any one of the proceeding aspects wherein the candidate targets or biomarkers of the disease are nucleic acid sequences or proteins related to the one or more associated molecular features.

[0191] Aspect 32. The method of aspect 31 wherein the one or more associated molecular features are nucleic acid sequences and / or proteins, and the candidate targets or biomarkers of the disease are the nucleic acid sequences and / or proteins.

[0192] Aspect 33. The method of any one of the proceeding aspects wherein the knowledge graph is restricted to the first cell type for which the molQTL data was obtained or to a group of cell types that contains the first cell type from which the molQTL data and cell types that have the capability to change into the first cell type. Aspect 34. A method of treating a subject suffering from a disease, wherein a key target identified by the method according to any one of the proceeding aspects is targeted by a therapy in order to treat or prevent the disease.

[0193] Aspect 35. A method for determining whether a subject has a disease, wherein the method comprises determining the presence of a key biomarker in a sample obtained from the subject, wherein the key biomarker is identified by the method according to any one of the proceeding aspects.

[0194] Aspect 36. A system comprising: one or more processors; and a memory storing instructions which, when executed by the one or more processors, cause the one or more processors to: obtain, by the one or more processors, data for genetic risk variants associated with a first feature of the disease; obtain, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature; determine, by the one or more processors, one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features; identify, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generate, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculate, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.

[0195] Aspect 37. A non-transitory computer-readable medium storing instructions which, when executed by a processing unit comprising one or more processors, cause the one or more processors to: obtain data for genetic risk variants associated with a first feature of the disease; obtain cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature; determine one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features; identify a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generate a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculate a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.

[0196] Aspect 38. A method of identifying key targets or biomarkers for a disease, the method comprising : obtaining, by one or more processors, data for genetic risk variants associated with a plurality of features of the disease; obtaining, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a plurality of cell types, wherein each of the molQTL is associated with a molecular feature; determining, by the one or more processors, a matrix of genetic variants having a causal relationship with one or more of the plurality of features of the disease and one or more associated molecular features for the plurality of cell types; identifying, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generating, by the one or more processors, a first cell typespecific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes is representing one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculating, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.

Claims

CLAIMS1. A method of identifying key targets or biomarkers for a disease, the method comprising: obtaining, by one or more processors, data for genetic risk variants associated with a first feature of the disease; obtaining, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature; determining, by the one or more processors, one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features; identifying, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generating, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculating, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.

2. The method of claim 1 further comprising: outputting, by the one or more processors, the first cell type-specific knowledge graph.

3. The method of claim 2 wherein the step of outputting the first cell type-specific knowledge graph comprises storing the first cell type-specific knowledge graph at a storage location of a computing device.41REPLACEMENT SHEETSUBSTITUTE SHEET (RULE 26)4. The method of claim 2 wherein the step of outputting the first cell type-specific knowledge graph comprises displaying a graphical representation of the first cell type-specific knowledge graph within a graphical user interface.

5. The method of claim 1 further comprising: identifying, by the one or more processors, a first candidate target or biomarker, or a first common interactor, based on the level of interconnectivity of a respective node representing the first candidate target or biomarker, or the first common interactor.

6. The method of claim 5 further comprising: outputting, by the one or more processors, the first candidate target or biomarker, or the first common interactor.

7. The method of claim 1 wherein the causal relationship is determined from the data for genetic risk variants and the molQTL data using mendelian randomization.

8. The method of claim 1, further comprising: obtaining, by the one or more processors, a second knowledge graph comprising a second plurality of nodes, wherein at least one node of the first plurality of nodes and at least one node of the second plurality of nodes are related.

9. The method of claim 8 wherein the second knowledge graph is associated with cell type-specific molecular quantitative trait loci (molQTL) data for a second cell type.

10. The method of claim 8 wherein the second knowledge graph is associated with data for genetic risk variants associated with a second feature of the disease.

11. The method of claim 8 further comprising: generating, by the one or more processors, a third knowledge graph comprising the first cell type-specific knowledge graph and the second knowledge graph, wherein the third knowledge graph comprises at least one edge connecting the at least one node of the first plurality of nodes and the at least one node of the second plurality of nodes,wherein a candidate target or biomarker, or a common interactor, represented by a node within the third knowledge graph having a high level of interconnectivity is identified as a key target or biomarker.

12. The method of claim 11 wherein each at least one edge connects nodes representing a related candidate target or biomarker, related molecular pathway, or related common interactor.

13. The method of claim 1 wherein the first cell type-specific knowledge graph is generated from a parent knowledge graph.

14. The method of claim 13 wherein the first cell type-specific knowledge graph is generated by pruning the parent knowledge graph in accordance with a locality constraint with respect to nodes within the parent knowledge graph representing the candidate targets or biomarkers.

15. The method of claim 14 wherein the locality constraint is a shortest path constraint such that a shortest path distance between pairs of nodes representing candidate targets or biomarkers within the first cell type-specific knowledge graph is less than a predetermined distance.

16. The method of claim 15 wherein the predetermined distance is less than 4.

17. The method of claim 1 wherein the level of interconnectivity of a node is one of: a degree of the node, or an eigencentrality of the node.

18. The method of claim 1 wherein a node is determined to have a high level of interconnectivity if the level of interconnectivity for the node is within an upper quartile of levels of interconnectivity for all nodes within the first cell type-specific knowledge graph.

19. The method of claim 1 wherein nodes within the first cell type-specific knowledge graph associated with molecular pathways include nodes representing gene expression pathways and / or nodes representing protein-protein interaction pathways.

20. The method of claim 1 wherein the key target is a target that can be targeted by a therapy in order to treat or prevent the disease.

21. The method of claim 20 wherein the key target is a nucleic acid sequence or a protein.

22. The method of claim 1 wherein the key biomarker for a disease is indicative of the presence or severity of the disease.

23. The method of claim 22 wherein the key biomarker is presence or level of a nucleic acid sequence or a protein, or the level of methylation or phosphorylation of a nucleic acid.

24. The method of claim 1 wherein the genetic risk variants associated with the first feature of the disease are: single-nucleotide polymorphisms / variations (SNPs / SNVs); insertion and / or deletion variations; copy number variations; translocation and / or an inversions, or a combination thereof.

25. The method of claim 24 wherein the genetic risk variants are SNPs.

26. The method of claim 1 wherein the disease is: heart disease, such as ischemic heart disease and coronary artery disease; cancer; a bacterial or viral infection; a cerebrovascular disease, such as stroke; a respiratory disease, such as chronic obstructive pulmonary disease; diabetes; an autoimmune disease, such as ulcerative colitis, Crohn's disease, inflammatory bowel disease, rheumatoid arthritis, Guillain-Barre syndrome, Sjogren's syndrome, scleroderma and Graves' disease; or a neurological disorder, such as Alzheimer's disease, multiple sclerosis, Parkinson's disease, amyotrophic lateral sclerosis (ALS) and frontotemporal dementia.

27. The method of claim 1 wherein the molQTL data associates a genetic variant with a molecular feature and the genetic variant is selected from the group consisting of: single-nucleotide polymorphisms / variations (SNPs / SNVs); insertion and / or deletion variations; copy number variations; translocation and / or an inversions, or a combination thereof.

28. The method of claim 27 wherein the genetic risk variant is a SNP.

29. The method of claim 1 wherein the molQTL data is cell type-specific expression quantitative trait loci (eQTL) data, cell type-specific DNA methylation quantitative trait loci (meQTL) data, cell type-specific histone modification quantitative trait loci(hQTL) data, cell type-specific chromatin accessibility quantitative trait loci (caQTL) data, cell type-specific alternative gene splicing quantitative trait loci (sQTL) data, cell type-specific protein level quantitative trait loci (pQTL) data, cell type-specific microRNA expression quantitative trait loci (mirQTL) data or cell type-specific ribosome occupancy quantitative trait loci (rQTL) data.

30. The method of claim 1 wherein the data for the one or more genetic risk variants is harmonized with the molQTL data prior to the step of determining the causal relationship.

31. The method of claim 1 wherein the candidate targets or biomarkers of the disease are nucleic acid sequences or proteins related to the one or more associated molecular features.

32. The method of claim 31 wherein the one or more associated molecular features are nucleic acid sequences and / or proteins, and the candidate targets or biomarkers of the disease are the nucleic acid sequences and / or proteins.

33. The method of claim 1 wherein the knowledge graph is restricted to the first cell type for which the molQTL data was obtained or to a group of cell types that contains the first cell type from which the molQTL data and cell types that have the capability to change into the first cell type.

34. The method of claim 1, wherein the key target as identified within the first cell type-specific knowledge graph is targeted by a therapy in order to treat or prevent a disease of a subject suffering from the disease.

35. The method of claim 1 further comprising determining the presence of a key biomarker in a sample obtained from a subject, wherein the key biomarker of the subject is the key biomarker is as identified within the first cell type-specific knowledge graph.

36. A system comprising: one or more processors; and a memory storing instructions which, when executed by the one or more processors, cause the one or more processors to: isobtain, by the one or more processors, data for genetic risk variants associated with a first feature of the disease; obtain, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature; determine, by the one or more processors, one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features; identify, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generate, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculate, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.

37. A non-transitory computer-readable medium storing instructions which, when executed by a processing unit comprising one or more processors, cause the one or more processors to: obtain data for genetic risk variants associated with a first feature of the disease; obtain cell type-specific molecular quantitative trait loci (molQTL) data for a first cell type, wherein each of the molQTL is associated with a molecular feature; determine one or more genetic risk variants having a causal relationship with the first feature of the disease and one or more associated molecular features;identify a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generate a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes represents one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculate a level of interconnectivity for each node within the first cell type-specific knowledge graph, wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.

38. A method of identifying key targets or biomarkers for a disease, the method comprising: obtaining, by one or more processors, data for genetic risk variants associated with a plurality of features of the disease; obtaining, by the one or more processors, cell type-specific molecular quantitative trait loci (molQTL) data for a plurality of cell types, wherein each of the molQTL is associated with a molecular feature; determining, by the one or more processors, a matrix of genetic variants having a causal relationship with one or more of the plurality of features of the disease and one or more associated molecular features for the plurality of cell types; identifying, by the one or more processors, a plurality of candidate targets or biomarkers of the disease based on the one or more associated molecular features; generating, by the one or more processors, a first cell type-specific knowledge graph comprising a first plurality of nodes, wherein each of the first plurality of nodes is representing one of the plurality of candidate targets or biomarkers, a molecular pathway associated with the candidate targets or biomarkers, or one or more common interactors; and calculating, by the one or more processors, a level of interconnectivity for each node within the first cell type-specific knowledge graph,wherein a candidate target or biomarker, or a common interactor, represented by a node having a high level of interconnectivity within the first cell type-specific knowledge graph is identified as a key target or biomarker.IB

Images