Identification of missing variants by combining multiple analytic pipelines.

Area of Science:

• Genomics
• Bioinformatics
• Human Genetics

Background:

• Genome-wide association studies (GWAS) traditionally identified risk factors.
• Complex disease research now focuses on sequencing-based rare variant discovery.
• Large sample sizes are required for statistical power in sequencing studies.

Purpose of the Study:

• To evaluate the adequacy of current variant calling practices for large sequencing cohorts.
• To assess the impact of using multiple analytic pipelines versus a single pipeline.
• To determine the proportion of high-quality variants missed by single-pipeline approaches.

Main Methods:

• Analyzed 10,000 exomes from the Alzheimer's Disease Sequencing Project (ADSP).
• Compared variants identified by different read aligners and variant calling strategies.
• Assessed the effect of combining multiple pipelines on variant recovery.

Main Results:

• Single pipelines missed an increasing number of high-quality variants with larger sample sizes.
• Combining two read aligners and two variant calling strategies rescued 30% of variants at 2000 samples and 56% at 10,000 samples.
• Rescued variants included a higher proportion of low-frequency and rare variants, including known pathogenic mutations in Alzheimer's disease genes.

Conclusions:

• Single bioinformatics pipeline approaches are inadequate for large-scale sequencing projects.
• Multi-pipeline strategies are essential for comprehensive variant set identification.
• Improved variant detection is critical for genetic association analyses of complex diseases.