Recurrent miscalling of missense variation from short-read genome sequence data.

Area of Science:

• Genomics
• Bioinformatics
• Population Genetics

Background:

• Short-read genome resequencing generates extensive genetic variation data.
• Exhaustive validation of numerous variants is often infeasible.
• Undetected variant miscalls can disproportionately impact individual genome interpretation and public variation databases.

Purpose of the Study:

• To investigate the nature and prevalence of sequence variation miscalling in short-read data.
• To identify sequence-intrinsic factors contributing to recurrent false positive variant calls.
• To develop methods for identifying and mitigating miscalled variants.

Main Methods:

• Analysis of short-read sequencing data from human exomes and mouse strains.
• Simulations involving read resampling, realignment, and variant recalling.
• Assessment of variant call sensitivity to sequence read length and genomic distance from reference.

Main Results:

• Recurrent, sequence-intrinsic miscalling of genetic variants is prevalent in short-read data.
• Miscalls are sensitive to read length and arise from alignment difficulties in redundant genomic regions.
• Thousands of recurrent false positive variants were identified per individual (human and mouse), many present in public databases.
• Over two-thirds of false positive variation can be identified using simulation-based approaches.

Conclusions:

• Removing recurrent false positives enhances the quality of individual variant datasets.
• Miscalled variants are sequence-specific and can be characteristic of individuals, pedigrees, or ethnic groups.
• Read length significantly influences variant calling, impacting cohort studies using diverse datasets.