VariFAST

VariFAST automates false-positive filtering and refinement of germline and somatic variants from whole-genome and exome next-generation sequencing (NGS) data to improve the accuracy and consistency of variant calls.

Key Features:

• Automated false-positive filtering: Processes BAM and VCF files to identify and filter false-positive variants.
• Variant scoring (v-score): Computes a variant score (v-score) derived from a weighted sum of metrics that contribute to false positives.
• Tagged-signature method: Implements the "Variant Filter by Automated Scoring based on Tagged-signature" approach for automated scoring.
• GATK integration: Integrates with the GATK Best Practices pipeline for incorporation into standard variant-calling workflows.
• Predictive modeling (XGBOOST): Includes a predictive model trained using the XGBOOST algorithm to enhance germline variant refinement.
• Benchmark validation: Validated on Genome in a Bottle Consortium (GIAB) benchmark datasets and on somatic variant datasets from malignant carcinomas and benign adenomas.
• Comparison to VQSR: Demonstrates superior performance compared to VQSR, particularly for INDEL filtering.
• Reduction of manual review variability: Automates the manual review step to reduce inter- and intra-laboratory variability in variant filtering.

Scientific Applications:

• Germline variant refinement: Refinement of germline SNV and INDEL calls from whole-genome and exome NGS data.
• Somatic variant filtering: Filtering of somatic variants in cancer datasets, including malignant carcinomas and benign adenomas.
• Benchmarking and method comparison: Method benchmarking using GIAB datasets and comparative evaluation against VQSR.
• Integration in variant-calling workflows: Deployment within GATK Best Practices pipelines for end-to-end variant detection and refinement.

Methodology:

Processes BAM and VCF files; implements the Variant Filter by Automated Scoring based on Tagged-signature method; computes a v-score as a weighted sum of false-positive metrics; integrates with the GATK Best Practices pipeline; and uses a predictive model trained with the XGBOOST algorithm.