VariantMetaCaller: automated fusion of variant calling pipelines for quantitative, precision-based filtering

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Low concordance among variant calling methods challenges next-generation sequencing (NGS) applications.
• Hard filtering and variant quality score recalibration have limitations, including threshold selection complexity and data requirements.

Purpose of the Study:

• To develop a novel method, VariantMetaCaller, for improving germline variant calling accuracy.
• To test the hypothesis that automated fusion of variant calling pipeline information enhances performance over existing methods.

Main Methods:

• Evaluated pipelines using BWA/Bowtie2 aligners with GATK, FreeBayes, and SAMtools.
• Introduced VariantMetaCaller, employing Support Vector Machines to integrate information from multiple variant callers.
• Assessed performance using simulated and real benchmark sequencing data (NA12878).

Main Results:

• VariantMetaCaller demonstrated significantly higher sensitivity and precision across various target region sizes (kilobases to whole exomes).
• The method effectively fuses complementary information from discordant variant callers.
• VariantMetaCaller enables quantitative, precision-based filtering and balances sensitivity and precision.

Conclusions:

• VariantMetaCaller is a viable alternative to hard filtering and recalibration, especially when high-quality variant call sets are unavailable.
• Applicable to both small target regions and whole exomes.
• Freely available at http://bioinformatics.mit.bme.hu/VariantMetaCaller.