#2 · Genomics

Context-aware models refine variant calling, eQTL mapping and polygenic prediction

A recurring thread is explicit modelling of the structure and context that shapes variant-level inference. Linked-read and short-read sequencing with a transmission-based framework detects and validates copy number variants in New Zealand dairy cattle. In prediction, RICE combines common-variant ensemble learning with rare-variant gene-level testing across diverse populations, IFAM integrates many functional annotation types into linear mixed models, and GxE-aware multi-kernel prediction targets ash dieback resistance across environments. Regulatory network topology, linking motifs, modules and hub regulators to cis- and trans-heritability of expression, and CASE, which jointly fine-maps single-cell eQTLs across cell types, extend this modelling to gene regulation. What is new is treating cell-type specificity, network architecture, annotation context and environment as explicit model components across detection and prediction. A polygenic score reproducibility study across 115 traits adds empirical analytic validity assessment.