Multiomics integration prioritizes potential drug targets for multiple sclerosis

Abstract

Multiple sclerosis (MS) is an immune-mediated disease with no current cure. Drug discovery and repurposing are essential to enhance treatment efficacy and safety. We utilized summary statistics for protein quantitative trait loci (pQTL) of 2,004 plasma and 1,443 brain proteins, a genome-wide association study of MS susceptibility with 14,802 cases and 26,703 controls, both bulk and cell-type specific transcriptome data, and external pQTL data in blood and brain. Our integrative analysis included a proteome-wide association study to identify MS-associated proteins, followed by summary-data-based Mendelian randomization to determine potential causal associations. We used the HEIDI test and Bayesian colocalization analysis to distinguish pleiotropy from linkage. Proteins passing all analyses were prioritized as potential drug targets. We further conducted pathway annotations and protein-protein interaction network analysis (PPI) and verified our findings at mRNA and protein levels. We tested hundreds of MS-associated proteins and confirmed 18 potential causal proteins (nine in plasma and nine in brain). Among these, we found 78 annotated pathways and 16 existing non-MS drugs targeting six proteins. We also identified intricateAQ PPIs among seven potential drug targets and 19 existing MS drug targets, as well as PPIs of four targets across plasma and brain. We identified two targets using bulk mRNA expression data and four targets expressed in MS-related cell types. We finally verified 10 targets using external pQTL data. We prioritized 18 potential drug targets in plasma and brain, elucidating the underlying pathology and providing evidence for potential drug discovery and repurposing in MS.

Keywords: causal; genomics; multiple sclerosis; proteomics; transcriptomics.

Fig. 1.

Flowchart of the overall study design. To identify proteins as potential drug targets for multiple sclerosis (MS) treatment, we first conducted a proteome-wide association study (PWAS) to detect proteins associated with MS susceptibility. For candidate proteins identified by PWAS, we conducted a summary-data-based Mendelian randomization (SMR) to test for pleiotropic associations between protein levels and MS. We also performed the HEIDI test and Bayesian colocalization analysis to distinguish pleiotropy from genetic linkage. Proteins that passed all three analyses (SMR analysis + HEIDI test + colocalization analysis) were identified as potential causal proteins. For all detected potential causal proteins, we conducted protein pathway annotation, identified drugs targeting these proteins, and analyzed protein–protein interactions among themselves, as well as interactions with protein targets of existing MS drugs. To gain additional layers of insight, we further investigated mRNA levels of these proteins to assess adherence to the central dogma of molecular biology (genetic locus→mRNA transcription→protein translation→MS) through SMR. Additionally, we investigated whether the genes encoding prioritized proteins are expressed in MS-related cell types through differential expression analysis on single-cell transcriptomes. Finally, we verified our prioritized proteins using external pQTL data through the Mendelian randomization framework. pQTL: protein quantitative trait loci; MS: multiple sclerosis; HEIDI: heterogeneity in dependent instruments test; eQTL: expression quantitative trait loci. GWAS: genome-wide association study.

Fig. 2.

Identifying candidate proteins associated with multiple sclerosis by integrating pQTL and GWAS through PWAS (A) Identifying candidate proteins in plasma associated with multiple sclerosis by integrating pQTL and GWAS through PWAS. (B) Identifying candidate proteins in brain associated with multiple sclerosis by integrating pQTL and GWAS through PWAS. The x-axis represents chromosomes. The y-axis represents the negative logarithm of P-values. Each dot on the Manhattan plot represents a gene, whose cis-regulated protein expression level was tested in association with the multiple sclerosis risk. The red dashed line indicates a nominal P-value threshold of 0.05, and the blue dashed line indicates the Bonferroni-corrected P-value threshold of 6.63 × 10⁻⁵ in plasma and 2.86 × 10⁻⁵ in brain. Proteins exhibiting nominal significance in association with multiple sclerosis risk were identified as candidate proteins, and their corresponding encoding genes are highlighted with red dots. pQTL: protein quantitative trait loci; GWAS: genome-wide association study; PWAS: proteome-wide association study; Chr: chromosome.

Fig. 3.

Identifying potential causal proteins for multiple sclerosis through SMR and colocalization analysis, based on candidate proteins determined by PWAS (A) Identifying potential causal proteins in plasma for multiple sclerosis through SMR and colocalization analysis, based on candidate proteins determined by PWAS. (B) Identifying potential causal proteins in brain for multiple sclerosis through SMR and colocalization analysis, based on candidate proteins determined by PWAS. The corresponding encoding genes for candidate proteins are listed outside the circle. Odds ratios from SMR analyses are presented in the outer circle. P-values for HEIDI tests are displayed in the middle circle. The posterior probabilities for colocalization analysis, under the hypotheses of one shared SNP associated with both protein quantitative traits and multiple sclerosis, are shown in the inner circle. The color of boxes in the circles represents the magnitude of each statistic, with warmer colors indicating stronger effects and gray indicating missing values. Significant results are highlighted with red asterisks, defined as Benjamini–Hochberg false discovery rate–adjusted P < 0.05 for SMR analysis, P_HEIDI > 0.05, and posterior probability for colocalization analysis > 0.75. Gene names in red indicate that their corresponding proteins passing all three analyses are identified as potential causal proteins. Gene names in brown indicate that their corresponding proteins are potential causal proteins and adhere to the central dogma of molecular biology. Fig. 3B excludes 127 out of 212 PWAS-identified candidate proteins from SMR analysis because their pQTLs had P-values > 5 × 10⁻⁸, not meeting the SMR instrument variable criterion (pQTLs with P-values < 5 × 10⁻⁸). The figure illustrates SMR and colocalization results for the remaining 85 out of 212 candidate proteins. SMR: summary-data-based Mendelian randomization; PWAS: proteome-wide association study; OR: odds ratio; HEIDI: heterogeneity in dependent instruments test. PP.H4: posterior probability for colocalization analysis under the hypotheses of one shared SNP associated with both traits.

Fig. 4.

Potential drug targets, protein-targeted drugs, and protein–protein interaction network. The corresponding encoding genes for these potential causal proteins are presented. Gene names in red indicate that their corresponding proteins adhere to the central dogma of molecular biology. Potential drug targets highlighted in yellow boxes are identified in plasma, and those in green boxes are identified in brain. Drug relations refer to the existing drugs targeting the identified potential drug targets. MS drug targets refer to proteins targeted by existing MS drugs. The protein–protein interaction network is presented between potential drug targets and current MS drug targets, with lines of different colors representing various types of interaction. MS: multiple sclerosis; HEIDI: heterogeneity in dependent instruments test.

Fig. 5.

Protein–protein interaction network analysis among all identified potential drug targets in plasma and brain. Protein–protein interaction network analysis was conducted among all detected potential drug targets (nine in plasma and nine in brain). Only those with a minimum required interaction score of 0.4 (six in plasma and five in brain) are considered interactions and presented in this figure. The corresponding encoding genes for these potential causal proteins are presented. Potential drug targets in yellow boxes are identified in plasma, and those in green boxes are identified in brain. The different colored lines connecting these potential drug targets represent various types of interactions.