#25 · Genomics

Causal and network-based inference from genetic data to nominate candidate causal genes

The dominant thread is converting genetic and network data into prioritized or causal gene and variant candidates through statistical and computational frameworks rather than direct experimental perturbation. Causal ASCEND applies constraint-based discovery with ancestral conditioning sets to scale gene regulatory network inference to high-dimensional multi-omics data. NERINE integrates gene and protein interaction networks into a probabilistic rare variant burden test to increase power. K-PoPS reformulates PoPS as a kernelized method for GWAS effector gene prioritization and decomposes each prediction into training-gene contributions and anchor scores. A benchmark of five unsupervised minimum-rank integrations of GWAS, eQTL, pQTL and exome scores addresses recovery of approved drug targets across 30 complex traits. HetNetEX provides analytic null distributions for path counts in heterogeneous biomedical knowledge graphs, and RankVar applies random forests to rare disease variant ranking. What is new is scale, analytic inference, and explainability. Outputs remain statistical inferences and candidate rankings rather than validated mechanisms.