Reducing false-positive incidental findings with ensemble genotyping and logistic regression based variant filtering...

Area of Science:

• Genomics and Bioinformatics
• Genetic Variant Analysis

Background:

• Whole genome sequencing (WGS) identifies disease-associated variants but is prone to false positives from sequencing and variant calling errors.
• Orthogonal sequencing methods reduce false positives but are cost-prohibitive for routine application.

Purpose of the Study:

• To develop and evaluate variant filtering approaches using logistic regression (LR) and ensemble genotyping to minimize false positives in WGS data.
• To assess the sensitivity and specificity of these methods compared to traditional genotype quality score filtering.

Main Methods:

• Paired WGS datasets from an extended family using two sequencing platforms were utilized.
• Variant filtering was performed using logistic regression (LR) and ensemble genotyping.
• Performance was evaluated using a validated set of variants in NA12878 and de novo mutation (DNM) discovery.

Main Results:

• LR and ensemble genotyping significantly reduced false-negative rates (1.1- to 17.8-fold) at comparable false discovery rates for single nucleotide variants (SNVs), insertions, and deletions.
• Ensemble genotyping excluded over 98% of false positives while retaining over 95% of true positives in DNM discovery.
• The proposed methods outperformed a consensus method using two sequencing platforms.

Conclusions:

• Logistic regression and ensemble genotyping are effective in minimizing false positives in WGS data without compromising sensitivity.
• Ensemble genotyping is particularly effective for de novo mutation discovery, essential for prioritizing phenotype-associated variants.
• These methods offer a cost-effective alternative to orthogonal sequencing for improving variant accuracy.