A Statistical Framework to Identify Cell Types Whose Genetically Regulated Proportions are Associated with Complex Diseases

Abstract

Finding disease-relevant tissues and cell types can facilitate the identification and investigation of functional genes and variants. In particular, cell type proportions can serve as potential disease predictive biomarkers. Here, we introduce a novel statistical framework, cell-type Wide Association Study (cWAS), that integrates genetic data with transcriptomics data to identify cell types whose genetically regulated proportions (GRPs) are disease/trait-associated. On simulated and real GWAS data, cWAS showed substantial statistical power with newly identified significant GRP associations in disease-associated tissues. More specifically, GRPs of endothelial and myofibroblasts in lung tissue were associated with Idiopathic Pulmonary Fibrosis and Chronic Obstructive Pulmonary Disease, respectively. For breast cancer, the GRP of blood CD8+ T cells was negatively associated with breast cancer (BC) risk as well as survival. Overall, cWAS is a powerful tool to reveal cell types associated with complex diseases mediated by GRPs.

MiXcan: a framework for cell-type-aware transcriptome-wide association studies with an application to breast cancer

“…Recent studies have also explored methods for performing celltype-level association analyses when the tissue-level data are available for all study participants 19,[38][39][40][41][42] . Luo et al evaluated cell-type-specific associations between DNA methylation and traits 40 , but this method did not involve prediction models and methylation data are required for all subjects.…”

“…Luo et al evaluated cell-type-specific associations between DNA methylation and traits 40 , but this method did not involve prediction models and methylation data are required for all subjects. Liu et al built tissue-level GReX prediction models, inferred cell types from the predicted GReX, and looked for associations of the inferred cell-type proportions with disease rather than constructing a TWAS framework for identifying genes 42 . In this study, MiXcan enables cell-type-aware TWAS in large populations using existing GWAS datasets that do not have transcriptomic and cell composition data from the disease-relevant tissue.…”

Best practices for multi-ancestry, meta-analytic transcriptome-wide association studies: Lessons from the Global Biobank Meta-analysis Initiative

“…Recent methods that study the mediation of the SNP-trait relationship by cell-type heterogeneity show that cell types and cell states are influenced by genetics and predict complex traits, and modeling these directly may lead to improved power in detecting trait associations. 26 , 69 – 72 Single-cell eQTL datasets can be integrated with GWASs to identify context-specific expression pathways that are disease related. Incorporating single-cell expression data into a predictive model will require improved methodology that models cell identity as a spectrum.…”