BAYSIC: a Bayesian method for combining sets of genome variants with improved specificity and sensitivity

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate genomic variant detection is crucial for clinical applications of genome data.
• Low concordance among variant callers hinders the clinical validity of whole genome and exome sequencing.
• Existing methods for combining variant calls lack accuracy and require training data.

Purpose of the Study:

• To develop BAYSIC (BAYeSian Integrated Caller), a novel method for integrating SNP variant calls from diverse callers.
• To improve the accuracy, sensitivity, and specificity of variant detection.
• To provide a flexible tool for both germline and somatic variant analysis.

Main Methods:

• BAYSIC employs an unsupervised, fully Bayesian latent class analysis to estimate error rates for each input variant caller.
• It does not require a "gold standard" training dataset.
• Users can adjust a posterior probability threshold to balance sensitivity and specificity.

Main Results:

• BAYSIC demonstrated superior variant-calling accuracy compared to other methods in assessments.
• Performance was validated using low-coverage samples, tumor/normal exome pairs, and known clinically relevant SNPs.
• The method showed effectiveness in both somatic mutation and germline variant detection.

Conclusions:

• BAYSIC effectively combines SNP variant calls from multiple programs, enhancing overall accuracy.
• The integrated variant calls improve both sensitivity and specificity.
• BAYSIC is applicable for combining germline variants and somatic mutations in tumor/normal sequencing data.