Development of an evidence-based algorithm in WES-based diagnostics
Sensitivity of whole exome sequencing (WES) is not well-defined. We applied very low thresholds in WES-associated variant calling to also enable investigation of candidate variants that are commonly neglected. As Sanger sequencing revealed ~5% of these to be true positives (Figure 1), we considered numerous variant-specific features (Tables 1 and 2) for the development of a robust predictor for true and false positives. Iterative rounds of receiver operating characteristic (ROC) curve generation identified features and corresponding thresholds with high predictive value (Figure 2). In a corresponding workflow for our data, 91.3% of variants can be pre-classified with 100% specificity and 99.8% sensitivity, while the remaining 8.7% of variants require confirmatory Sanger sequencing (Figure 3).