BATCAVE: calling somatic mutations with a tumor- and site-specific prior

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Detecting low-frequency somatic mutations is crucial for understanding tumor evolution, treatment resistance, and patient prognosis.
• Existing algorithms struggle to accurately identify these low-frequency variants due to their inherent biological characteristics.
• Tumors exhibit unique mutation profiles, a factor not adequately modeled by current variant detection methods.

Purpose of the Study:

• To develop a novel algorithm, BATCAVE (Bayesian Analysis Tools for Context-Aware Variant Evaluation), that incorporates tumor-specific mutational profiles.
• To improve the detection accuracy and reliability of low-frequency somatic mutations.
• To provide a computationally efficient and easily integrable tool for existing variant calling pipelines.

Main Methods:

• BATCAVE learns an individual tumor's mutational profile and mutation rate.
• These learned profiles are used as a prior in a Bayesian framework to evaluate potential mutations.
• An R implementation of BATCAVE was developed and integrated with the MuTect variant caller.

Main Results:

• Simulations demonstrated that BATCAVE significantly improves variant detection accuracy when added to MuTect.
• The algorithm enhances the calibration of posterior probabilities, allowing for better precision-recall trade-offs.
• BATCAVE also showed strong performance on real tumor genomic data.

Conclusions:

• BATCAVE offers a significant advancement in detecting low-frequency somatic mutations by contextualizing variant evaluation with tumor-specific biology.
• The R implementation is computationally efficient and readily adaptable to existing MuTect workflows.
• The BATCAVE framework can be extended to incorporate other biological features influencing mutation generation, broadening its applicability in cancer genomics.