CaDrA: A Computational Framework for Performing Candidate Driver Analyses Using Genomic Features.

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Identifying combinations of genetic alterations driving phenotypic outcomes is crucial for biological insight and therapeutic target discovery.
• Existing methods lack flexibility for joint analysis of multiple genomic alteration types and rigorous testing.

Purpose of the Study:

• To develop an integrative framework, Candidate Driver Analysis (CaDrA), for identifying functionally relevant subsets of genomic features associated with specific outcomes.
• To address limitations in analytical flexibility and significance testing of existing driver identification methods.

Main Methods:

• Developed CaDrA, an integrative framework using a step-wise heuristic search approach.
• Applied CaDrA to simulated multi-omic datasets, Cancer Cell Line Encyclopedia (CCLE) data, and The Cancer Genome Atlas (TCGA) data.
• Validated findings using in vitro experiments and assessed reproducibility with pan-cancer TCGA protein expression data.

Main Results:

• CaDrA demonstrated high sensitivity and specificity on simulated multi-omic data.
• Identified known mutations associated with drug sensitivity in cancer cell lines (CCLE).
• Discovered novel regulators of oncogenic activity (YAP/TAZ) in breast cancer (TCGA) and showed high reproducibility in pan-cancer analyses.

Conclusions:

• CaDrA is a powerful and flexible framework for querying large multi-omics datasets.
• Facilitates the discovery of potential drivers for various biological outcomes and therapeutic targets.
• Enables fast analysis of publicly available datasets like TCGA and CCLE.