Mergeomics 2.0: a web server for multi-omics data integration to elucidate disease networks and predict therapeutics

Abstract

The Mergeomics web server is a flexible online tool for multi-omics data integration to derive biological pathways, networks, and key drivers important to disease pathogenesis and is based on the open source Mergeomics R package. The web server takes summary statistics of multi-omics disease association studies (GWAS, EWAS, TWAS, PWAS, etc.) as input and features four functions: Marker Dependency Filtering (MDF) to correct for known dependency between omics markers, Marker Set Enrichment Analysis (MSEA) to detect disease relevant biological processes, Meta-MSEA to examine the consistency of biological processes informed by various omics datasets, and Key Driver Analysis (KDA) to identify essential regulators of disease-associated pathways and networks. The web server has been extensively updated and streamlined in version 2.0 including an overhauled user interface, improved tutorials and results interpretation for each analytical step, inclusion of numerous disease GWAS, functional genomics datasets, and molecular networks to allow for comprehensive omics integrations, increased functionality to decrease user workload, and increased flexibility to cater to user-specific needs. Finally, we have incorporated our newly developed drug repositioning pipeline PharmOmics for prediction of potential drugs targeting disease processes that were identified by Mergeomics. Mergeomics is freely accessible at http://mergeomics.research.idre.ucla.edu and does not require login.

Graphical Abstract

Mergeomics uses single omics or multi-omics data to produce pathway- and network-level mechanistic understanding of disease and identify potential therapeutic targets.

Figure 1.

Workflow of Mergeomics. We provide four options on the web server to tailor to the user's data type. Case One: Individual GWAS analysis. For GWAS datasets we advise utilizing the MDF function; however, we also provide the ability to skip MDF and directly run MSEA and follow the workflow to PharmOmics or KDA. Case Two: Individual EWAS, TWAS, PWAS or MWAS analysis. In this case, we directly start at MSEA without MDF; however, we also provide the ability to utilize the MDF function if needed. From here the user can feed the MSEA results into PharmOmics or KDA. Case Three: Multi-omics analysis. If the user has multiple omics of the same type (e.g. two GWAS) or different types (e.g. TWAS and EWAS), they can utilize the Meta-MSEA function to derive disease-associated pathways and can input their results into PharmOmics or KDA. Case Four: A gene list(s) to run KDA. The user in this case can upload their gene sets of interest and upload or select a network to derive KD genes and visualize top KD subnetworks. The disease subnetwork or significant KDs can be fed into PharmOmics for drug repositioning.

Figure 2.

Mergeomics pipeline inputs. MDF is the default starting point for GWAS analysis and is an optional step for EWAS/TWAS/PWAS/MWAS. MDF requires marker-disease associations, a marker-gene mapping file, and a marker dependency file. Users with GWAS data can also skip MDF and run MSEA directly. MDF produces corrected marker-disease associations and marker-gene mapping files containing independent markers that are used for MSEA. For MSEA, required files for all datasets are the marker-disease associations and marker sets (pathway/modules). The marker to gene mapping file is required for GWAS and EWAS and optional for MWAS, TWAS and PWAS. Disease-associated marker sets from MSEA can be fed into KDA, which requires gene sets and a network. KDA can also be a starting point of analysis. Disease-associated gene sets from MSEA or KDs and disease subnetwork from KDA can be fed into PharmOmics for drug repositioning.

Figure 3.

Top KDs network visualization. Screenshot of the in-browser interactive network visualization (using Cytoscape.js) directed from the KDA results page. The colors of the nodes represent member genes of a disease-associated pathway. The diamond shaped nodes represent KD genes, where the border color represents the top pathway that is regulated by the KD. If a node has multiple colors, it is part of two or more disease-associated pathways, and if a node is grey, it does not belong to the disease pathways (non-member genes) but is present in the input network.

Figure 4.

Meta-MSEA use case study overview. To showcase the function and output of the web server, we utilized multiple human psoriasis GWAS and EWAS data and ran the multiple omics data workflow (Case 3 in Figure 1, Meta-MSEA). Firstly, we uploaded the psoriasis GWAS data, mapped the SNPs to genes using a combined skin and blood eQTL file, and filtered for LD > 0.7 to remove redundant SNPs in LD. Next, we uploaded our psoriasis EWAS association datasets and mapped the CpG sites to genes based on a 5 kb distance. Finally, we uploaded KEGG pathways with a psoriasis control set. Pathway enrichment results are produced, and each pathway's top genes, markers, and corresponding association values are displayed. Psoriasis-associated pathways are used as input into KDA as well as PharmOmics drug repositioning (using genes from significant pathways/modules). In the KDA, along with the Meta-MSEA input, we chose the blood GIANT network option and ran the KDA providing KD results and visualization (Figure 3) and additionally utilized the network genes as an input into PharmOmics. Finally, two sets of drug repositioning results were produced using gene overlap-based drug repositioning in PharmOmics: one based on the genes of significant pathways from the Meta-MSEA results and the other based on the KDA subnetwork genes.

Figure 5.

Output files from Meta-MSEA, KDA, and PharmOmics based on the case study of psoriasis outlined in Figure 4. Tables are interactive with pagination, search, and sort functions. Result files are downloadable from links on the webpage above the output tables (not shown). (A) Example Meta-MSEA output from the psoriasis use case. The table shown details the significance of association of each pathway/module and the top markers and corresponding association strengths that contributed to the module association. There are two additional tables which can be displayed by clicking on the tabs to the right of ‘Module Results’ at the top. The second table shows the significance and details of merged modules after merging redundant pathways (termed ‘Supersets’), and these non-overlapping gene sets are used as input to KDA. The third table shows the individual significance values for each omics dataset included in this Meta-MSEA of one GWAS and two EWAS of psoriasis. (B) Example KDA output from the psoriasis use case. The table shown records the significance of KDs, the pathways/modules that they regulate based on network topology, and details of the local subnetwork such as the number of KD subnetwork genes and number of pathway/module gene overlap with the KD subnetwork. Merged pathways/modules are represented by the term ‘Superset’, which means they are comprised of multiple redundant (significant gene overlap) pathways. (C) Example PharmOmics drug repositioning output using a gene overlap-based analysis between disease pathways and drug signatures. Gene overlap-based drug repositioning queries all tissue- and species-specific meta-analyzed and dose/time segregated gene signatures of drugs in our PharmOmics database as well as all L1000 drug signatures. The table shown gives the dataset source of the drug signature, the method of differential gene expression analysis, details of the drug study including species, tissue or cell line, whether the study was done in vitro or in vivo, the dose and time regimen, the Jaccard score, and statistical significance of the gene overlap between the input psoriasis related genes from Meta-MSEA and the drug signatures.