Cornell, E. | bioRxiv Science

bioRxiv2022

Uncovering Cross-Cohort Molecular Features with Multi-Omics Integration Analysis

Integrative approaches that simultaneously model multi-omics data have gained increasing popularity because they provide holistic system biology views of multiple or all components in a biological system of interest. Canonical correlation analysis (CCA) is a correlation-based integrative method. It was initially designed to extract latent features shared between two assays by finding the linear combinations of features - referred to as canonical vectors (CVs) - within each assay that achieve maximal across-assay correlation. Sparse multiple CCA (SMCCA), a widely-used derivative of CCA, allows more than two assays but can result in non-orthogonal CVs when applied to high-dimensional data. Here, we incorporated a variation of the Gram-Schmidt (GS) algorithm with SMCCA to improve orthogonality among CVs. Applying our SMCCA-GS method to proteomics and methylomics data from the Multi-Ethnic Study of Atherosclerosis (MESA) and Jackson Heart Study (JHS), we identified strong associations between blood cell counts and protein abundance. This finding suggests that adjustment of blood cell composition should be considered in protein-based association studies. Importantly, CVs obtained from two independent cohorts demonstrate transferability across the cohorts. For example, proteomic CVs learned from JHS explain similar amounts of blood cell count phenotypic variance in MESA, explaining 39.0% ~ 50.0% variation in JHS and 38.9% ~ 49.1% in MESA, similar transferability was observed for other omics-CV-trait pairs. This suggests that biologically meaningful and cohort-agnostic variation is captured by CVs. We further developed Sparse Supervised Multiple CCA (SSMCCA) to allow supervised integration analysis for more than two assays. We anticipate that applying our SMCCA-GS and SSMCCA on various cohorts would help identify cohort-agnostic biologically meaningful relationships between multi-omics data and phenotypic traits. Author SummaryComprehensive understanding of human complex traits may benefit from incorporation of molecular features from multiple biological layers such as genome, epigenome, transcriptome, proteome, and metabolome. CCA is a correlation-based method for multi-omics data which reduces the dimension of each omic assay to several orthogonal components - commonly referred to as canonical vectors (CVs). The widely-used SMCCA method allows effective dimension reduction and integration of multi-omics data, but suffers from potentially highly correlated CVs when applied to high-dimensional omics data. Here, we improve the statistical independence among the CVs by adopting a variation of the GS algorithm. We applied our SMCCA-GS method to proteomic and methylomic data from two cohort studies, MESA and JHS. Our results reveal a pronounced effect of blood cell counts on protein abundance, strongly suggesting blood cell composition adjustment in protein-based association studies may be necessary. Finally, we present SSMCCA which allows supervised CCA analysis for the association between one phenotype of interest and more than two assays. We anticipate that SMCCA-GS would help reveal meaningful system-level factors from biological processes involving features from multiple assays; and SSMCCA would further empower interrogation of these factors for phenotypic traits related to health and diseases.

genomics↗

bioRxiv2021

Protein prediction for trait mapping in diverse populations

Genetically regulated gene expression has helped elucidate the biological mechanisms underlying complex traits. Improved high-throughput technology allows similar interrogation of the genetically regulated proteome for understanding complex trait mechanisms. Here, we used the Trans-omics for Precision Medicine (TOPMed) Multi-omics pilot study, which comprises data from Multi-Ethnic Study of Atherosclerosis (MESA), to optimize genetic predictors of the plasma proteome for genetically regulated proteome-wide association studies (PWAS) in diverse populations. We built predictive models for protein abundances using data collected in TOPMed MESA, for which we have measured 1,305 proteins by a SOMAscan assay. We compared predictive models built via elastic net regression to models integrating posterior inclusion probabilities estimated by fine-mapping SNPs prior to elastic net. In order to investigate the transferability of predictive models across ancestries, we built protein prediction models in all four of the TOPMed MESA populations, African American (n=183), Chinese (n=71), European (n=416), and Hispanic/Latino (n=301), as well as in all populations combined. As expected, fine-mapping produced more significant protein prediction models, especially in African ancestries populations, potentially increasing opportunity for discovery. When we tested our TOPMed MESA models in the independent European INTERVAL study, fine-mapping improved cross-ancestries prediction for some proteins. Using GWAS summary statistics from the Population Architecture using Genomics and Epidemiology (PAGE) study, which comprises ~50,000 Hispanic/Latinos, African Americans, Asians, Native Hawaiians, and Native Americans, we applied S-PrediXcan to perform PWAS for 28 complex traits. The most protein-trait associations were discovered, colocalized, and replicated in large independent GWAS using proteome prediction model training populations with similar ancestries to PAGE. At current training population sample sizes, performance between baseline and fine-mapped protein prediction models in PWAS was similar, highlighting the utility of elastic net. Our predictive models in diverse populations are publicly available for use in proteome mapping methods at https://doi.org/10.5281/zenodo.4837328. Author summaryGene regulation is a critical mechanism underlying complex traits. Transcriptome-wide association studies (TWAS) have helped elucidate potential mechanisms because each association connects a gene rather than a variant to the complex trait. Like genome-wide association studies (GWAS), most TWAS are still conducted exclusively in populations of European ancestry, which misses the opportunity to test the full spectrum of human genetic variation for associations with complex traits. Here, move beyond the transcriptome and because protein measurement assays are growing to allow interrogation of the proteome, we use data from TOPMed MESA to develop genetic predictors of protein abundance in diverse ancestry populations. We compare model-building strategies with the goal of providing the best resource for protein association discovery with available data. We demonstrate how these prediction models can be used to perform proteome-wide association studies (PWAS) in diverse populations. We show the most protein-trait associations were discovered, colocalized, and replicated in independent cohorts using proteome prediction model training populations with similar ancestries to individuals in the GWAS. We shared our protein prediction models and performance statistics publicly to facilitate future proteome mapping studies in diverse populations.

genomics↗