DriverSub

DriverSub predicts subgroup-specific driver genes from heterogeneous cancer mutation data using a subspace learning framework when subgroup annotations are unavailable.

Key Features:

• Subspace Learning Framework: Implements a subspace learning approach to infer subgroup-specific driver genes and implicit subgroup annotations from heterogeneous cancer samples.
• Precision Medicine Application: Targets identification of subgroup-specific driver genes to inform personalized treatment strategies.
• Robust Performance: Outperforms existing methods in predicting driver genes and associated subgroups on simulation datasets with known ground truth and on real mutation data.
• Validation Against Known Data: Shows high enrichment for experimentally validated driver genes and inferred subgroups that align significantly with annotated molecular subgroups in real cancer mutation datasets.

Scientific Applications:

• Tumorigenesis Research: Enables study of genetic drivers underlying tumorigenesis by identifying subgroup-specific driver genes in heterogeneous cancers.
• Precision Medicine: Facilitates development of subgroup-targeted therapeutic strategies by linking driver genes to specific patient subgroups without prior annotations.

Methodology:

DriverSub processes raw mutation data using the script ./GenerateInputData.m, which calls ./bin/P02_GenerateMutData.m to generate mutation matrices for cancer types such as BRCA and BLCA; these matrices serve as input data for DriverSub's analysis.