De novo identification of complex multimorbid conditions by integration of gene regulation and protein interaction networks with genome-wide association studies

Abstract

Network analysis represents a powerful approach for untangling the relationships between complex diseases. Here, we integrated information on physical contacts between common single nucleotide polymorphisms (SNPs; MAF≥0.05) and expressed genes (TPM>0.1), with expression quantitative trait loci (eQTL) data from whole blood to construct a blood-specific spatial gene regulatory network (GRN). Using this GRN, we located the genes that are functionally affected by asthma-associated SNPs to identify the asthma-specific module. We then expanded outwards to identify curated protein-protein interactions occurring 1 to 4 edges away from the asthma module. The eQTLs spatially regulating the identified genes were used to interrogate the GWAS Catalog to identify enriched traits (hypergeometric test; FDR≤0.05) within each of the 4 levels. This led to de novo identification of traits both previously reported (e.g., sarcoidosis and lung cancer) and unreported (e.g., disc-degeneration) to be multimorbid with asthma, and pinpoints the variants and genes bridging asthma to the multimorbidities. Our discovery approach identifies significantly enriched traits in the regulatory space proximal to a disease of interest, in the tissue of interest, without a priori selection of the interacting diseases. The predictions it makes expand our understanding of possible shared therapeutic targets, especially for diseases such as asthma where no cure is currently available.

De novo discovery of traits co-occurring with chronic obstructive pulmonary disease

“…The Multimorbid3D pipeline was used to identify traits that were co-occuring with COPD (Fig. 1B) (Zaied et al 2022). Briefly, 'all' eQTLs that were associated with the proteins encoded by the genes targeted by the COPD-eQTLs (level 0) were identified within the LSPPIN.…”

“…However, gene regulation is widely understood to occur through the combinatorial action of regulatory elements, transcription factors and genes within complex networks (i.e. gene regulatory network, GRN) (Buenrostro et al 2018; Chen et al 2021a; Zaied et al 2022; Chen et al 2021b). Moreover, genes encode proteins, that physically interact with each other to form a complex protein-protein interaction network (PPIN) that responds to biological and environmental signals.…”

Comorbidity genetic risk and pathways impact SARS-CoV-2 infection outcomes

“…Genes targeted by spatially constrained eQTLs were identified (FDR ≤ 0.05) within whole blood and brain cortex using 5,594 SNPs that were associated with SARS-CoV-2 hospitalisation or severe phenotype (Supplementary Figure 1c; Supplementary Table 6). GRNs for blood 36 and brain cortex (1,050,155 spatial eQTLs involving 862,964 SNPs and 14,428 genes; Supplementary Table 7) were generated.…”

“…Supplementary Figure 1c; Supplementary Table 6) targeted by SARS-CoV-2 associated eQTLs in these tissues (Supplementary Table 6), blood 36 and brain GRNs (Supplementary Table 7). Blood tissue traits identified from the STRING protein interactions that were (a) shared across both severe and hospitalised phenotypes; and (b) were shared and observed in the brain.…”

De novo discovery of traits co-occurring with chronic obstructive pulmonary disease

“…Little is known about functional relationships between genes and phenotypes in the lung. However, gene regulation is widely understood to occur through the combinatorial action of regulatory elements, transcription factors and genes within complex networks (i.e., gene regulatory network [GRN]) ( Buenrostro et al, 2018 ; Chen et al, 2021a , 2021b ; Zaied et al, 2022 Preprint ). Moreover, genes encode proteins that physically interact with each other to form a complex protein–protein interaction network (PPIN) that responds to biological and environmental signals.…”