CELL-TYPE DECONVOLUTION AND CANDIDATE GENE PRIORITIZATION IN CARDIAC COMPLEX TRAIT GENETICS

Bulk RNA-seq from heterogeneous tissues conflates two sources of variation: changes in celltypeabundance and changes in per-cell transcriptional regulation. In cardiac tissue, where cellularremodeling is both a hallmark and a driver of disease progression, this conflation can produce misleadingdifferential expression results and obscure the genetic architecture of complex traits. Thisdissertation develops and applies a composition-aware analytical framework for cardiac geneticsin the Collaborative Cross (CC), a genetically diverse mouse reference population. I first established a deconvolution and regression pipeline that partitions bulk expression intocompositional and molecular components. Applied to post-myocardial infarction hearts fromcardiomyocyte-specific α1A-adrenergic receptor knockout mice, this framework revealed that manytreatment-associated genes reflected shifts in fibroblast and immune cell abundance rather than intracellularregulatory changes (Chapter 2). To address the complementary challenge of prioritizingcandidate genes from the results of such transcriptomic studies as well as from population-scaleQTL mapping, I developed LocusPackRat, an R package that aggregates genomic, transcriptomic,and phenotypic evidence into standardized, auditable reports (Chapter 3). I then applied both tools to 63 CC strains subjected to chronic β-adrenergic stress via isoproterenol.QTL mapping identified 49 genome-wide significant loci for cardiac physiology traits, andLocusPackRat guided the reduction of over 2,000 positional candidates to a shortlist from whichsiRNA knockdown validated Abcb10, Mrps5, and Lmod3 as regulators of cardiomyocyte hypertrophy(Chapter 4). Extending deconvolution to this population revealed that cardiac cell-type proportions are heritable traits (H2 = 0.35 to 0.78) whose genetic control diminishes under pharmacologicalstress. Composition adjustment unmasked additional differentially expressed genes rather thanremoving false positives, and deregulation scoring classified 72% of treatment-responsive genes asmolecular in origin (Chapter 5). Together, these results establish that cell-type composition is a geneticallyregulated, environmentally sensitive intermediate phenotype that must beaccounted for in cardiac transcriptomic studies, and that transparent,multi-evidence candidate gene prioritization can bridge the gap between broadassociation intervals and functional validation.