Performance characterization of PCR-free whole genome sequencing for clinical diagnosis

Purpose: To evaluate the performance of PCR-free whole genome sequencing (WGS) for clinical diagnosis, and thereby revealing how experimental parameters affect variant detection. Methods: All the 5 NA12878 samples were sequenced using MGISEQ-2000. NA12878 samples underwent WGS with differing DNA input and library preparation protocol (PCR-based versus PCR-free protocols for library preparation). The DP (depth of coverage) and GQ (genotype quality) of each sample were compared. We developed a systematic WGS pipeline for the analysis of down-sampling samples of the 5 NA12878 samples. The performance of each sample was measured for sensitivity, coverage of depth and breadth of coverage of disease-related genes and CNVs. Results: In general, NA12878-2 (PCR-free WGS) showed better DP and GQ distribution than NA12878-1 (PCR-based WGS). With a mean depth of ~40X, the sensitivity of homozyous and heterozygous SNPs of NA12878-2 showed higher sensitivity (>99.77% and > 99.82%) than NA12878-1, and positive predictive value (PPV) exceeded 99.98% and 99.07%. The sensitivity and PPV of homozygous and heterozygous indels for NA12878-2 (PCR-free WGS) showed great improvement than NA128878-1. The breadths of coverage for disease-related genes and CNVs are slightly better for samples with PCR-free library preparation protocol than the sample with PCR-based library preparation protocol. DNA input also influences the performance of variant detection in samples with PCR-free WGS. Conclusion: Different experimental parameters may affect variant detection for clinical WGS. Clinical scientists should know the range of sensitivity of variants for different methods of WGS, which would be useful when interpreting and delivering clinical reports.