Impact of post-alignment processing in variant discovery from whole exome data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• GATK Best Practices recommend post-alignment processing (local realignment, BQSR) for large-scale sequencing.
• The benefit of these computationally intensive steps with updated mappers/callers and new sequencing platforms is unclear.
• Impact in highly divergent genomic regions, like the HLA system, remains unknown.

Purpose of the Study:

• To comprehensively assess the impact of post-processing on variant calling accuracy.
• To evaluate the influence of different mappers, callers, sequencing depth, and divergence levels.
• To determine if post-processing offers benefits across various analytical pipelines.

Main Methods:

• Utilized simulated and NA12878 exome data for analysis.
• Tested five to six popular mappers combined with five variant callers.
• Focused on chromosome 6p21.3 (HLA region) for high-divergence analysis.

Main Results:

• Local realignment had minimal impact on SNP calling but improved INDEL calling in specific pipelines.
• Base Quality Score Recalibration (BQSR) showed negligible effects on INDEL calling and often reduced SNP calling sensitivity.
• BQSR's impact on SNP calling varied by caller, coverage, and divergence, with reduced sensitivity in high-divergence regions.

Conclusions:

• The utility of post-processing is not universal and is highly dependent on the specific bioinformatics pipeline.
• Sequencing depth and the level of genomic divergence significantly influence the effectiveness of post-processing.
• Careful consideration of these factors is crucial when deciding whether to implement computationally intensive post-processing steps for exome data.