#5 · Tools & infrastructure

Computational models link genetic variants to RNA binding and cell-type expression

The dominant thread is computational inference of how genetic variation or molecular state maps to function and disease, using large multi-omics and single-cell resources. Within this, RNA-binding protein work is prominent: an expanded eCLIP resource covering 286 proteins in K562 and HepG2 supports deep-learning variant scoring and constraint estimates at binding sites, and MuSIC extends cross-species RBP-RNA prediction using RBP conservation across 11 species. CASE adds joint cell-type-specific and shared single-cell eQTL fine-mapping, reported to outperform SuSiE and mvSuSiE in simulation and OneK1K data. Multivariate genetic modeling resolves musculoskeletal disorders into three factors across 795 risk loci. Other items, including lncRNA prediction, driver-gene inference and survival mediation, apply related modeling to non-variant settings. Comparative claims are limited to the reported benchmarks.