Depth ratio distortions in whole exome sequencing (WES) as a quality control metric for copy number alterations calling
Authors
- BostonGene, Corp., Waltham, USA
Abstract
Changes in DNA, such as copy-number alterations (CNAs), are associated with many human cancers and may serve as predictive and prognostic biomarkers. However, depth ratio (DR) distortions, which are deviations in sequencing depth ratios between experimental samples and reference samples in WES, lead to false results for CNA calling. To address these challenges, we developed a distortion metric to capture repetitive DR deviations. It may serve as an additional QC metric that can enhance the reliability of WES analysis for the detection of CNAs.
METHODS
To develop the DR distortion metric, WES was performed on an internal cohort of 2,026 normal tissue samples. DR values were calculated by dividing the depth of experimental samples by the mean depth of reference samples (accounting for GC content and coverage bias). Sixteen samples with the most pronounced DR distortions (≥25%) were manually selected. Among all 1-Mb genome regions, the ones with the highest standard deviation of DR formed two clearly distinguishable clusters of regions with strong mutual anti-correlation, which were further fine-tuned using a pairwise correlation analysis. In each cluster, a mean DR was calculated within each chromosome arm. Next, a logarithmic ratio of DR means was obtained for 12 chromosome arms. Finally, the averaged value among the 12 arms was calculated as the final DR distortion metric, which was applied to all samples. Spearman's Rank Correlation Coefficients were calculated to assess correlation of our metric with standard QC metrics (e.g., mean GC content, coverage uniformity, and inner distance).
RESULTS
The developed DR distortion metric identified consistent patterns of distortion in 12 chromosome arms, often near telomeres and centromeres. Importantly, clinically significant genes (e.g., PTEN, MDM2, ERBB2, BRCA1, TSC1) were located within these regions. In 25 out of 2,026 samples DR increased by at least 20% in 3.6% of genomic regions and decreased by at least 20% in 9.3% of regions. Interestingly, our DR distortion metric showed strong correlations with mean GC content (ρ=0.85, P<0.001), coverage uniformity (ρ=0.36, P<0.001), and inner distance (ρ=-0.46, P<0.001), simultaneously revealing samples not detected by standard QC metrics.
CONCLUSIONS
Our developed DR distortion metric revealed consistent distortion patterns for DR values in normal samples, indicating that this new metric enhances the reliability of CNA detection in WES and can be generalized for use in tumor sample analysis.
Latest publications
