#38 · ML / AI methods

Somatic variants joined to dependency and single-cell context for target nomination

The dominant thread is joining somatic variant information with functional dependency or network context to prioritise cancer targets. CRISPR-VUS applies Cancer Dependency Map data to dependency-associated variants, nominating candidate drivers and repositioning opportunities. The newer elements move to single-cell and evolutionary resolution: SCANBIT calls SNVs on pseudobulked scRNA-seq clusters to identify tumour cell populations, and SigDyn couples variant detection, phylogeny inference and signature identification. Protein dosage buffering of aneuploidy is resolved through machine learning on copy number-dependent features, and an ensemble model on LGG expression with metabolic modelling defines survival risk groups. These extend the logic toward mechanism and progression rather than directly toward target lists.