tarSVM: Improving the accuracy of variant calls derived from microfluidic PCR-based targeted next generation...

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Targeted sequencing combined with microfluidic PCR and next-generation sequencing is cost-effective for screening monogenic diseases.
• PCR amplification in this pipeline introduces duplicate reads, challenging accurate variant calling.
• Existing methods using threshold-based filtering and manual inspection often fail Sanger sequencing validation.

Purpose of the Study:

• To design an improved variant filtering strategy that specifically addresses microfluidic PCR-associated challenges.
• To enhance the accuracy of variant calling in targeted sequencing data.

Main Methods:

• Developed an open-source variant filtering pipeline named tarSVM (targeted sequencing support vector machine).
• Utilized a Support Vector Machine (SVM) and a novel normalized allele dosage test.
• Incorporated variant features from Genome Analysis Toolkit and allele dosage information, trained on 1000 Genomes and ExAC data.

Main Results:

• tarSVM demonstrated higher accuracy in predicting Sanger sequencing validation compared to existing methods across two cohorts (84.5% vs 78.8% and 73.3% vs 61.5%).
• Achieved a false discovery rate of 5% on a validation cohort.
• The method effectively synthesizes variant features and allele dosage information.

Conclusions:

• tarSVM significantly increases variant calling accuracy for microfluidic PCR-based targeted sequencing.
• Leads to more confident downstream analyses and reduces costs associated with Sanger validation.
• Provides an open-source, less labor-intensive pipeline for the entire sequencing data analysis workflow.