(A) systematic comparison of whole genome sequencing and targeted panel sequencing for precision oncology

Abstract

The analysis of cancer genome provides clinically meaningful information such as druggable mutations. Whole genome sequencing (WGS) provides a complete mutational landscape at an affordable cost thereby having potential of increased clinical utility compared to the usual approach of targeted panel sequencing (TPS). To understand the capabilities of WGS in the precision oncology, I analyzed WGS (tumor and blood) and TPS (tumor only) of 60 non-small cell lung cancers. In addition to driver events, I compared pattern-based features such as tumor mutational burden (TMB), mutational signature, and copy number variation (CNV). Across the 60 samples, WGS and TPS showed 100% concordance in detecting 51 classical driver mutations fusion oncogenes. WGS detected 94.9% of non-classical driver mutations detected by TPS and additionally found 42 more events (e.g. PRKN mutation). Among non-classical driver structural variations (SVs), 3 were detected from both methods (e.g. CDKN2A truncation), whereas WGS detected 7 more events (e.g. RAF1-DAAM1 fusion). As for TMB, TPS values were comparable to that of WGS (Pearson r > 0.95) but filtering strategies using unrelated TPS samples were required. With regard to detecting a specific mutational signature, TPS showed 100% positive predictive value for tobacco-related and 100% sensitivity for APOBEC-related signature. For CNV, TPS missed 4 of 34 driver amplifications within TPS target regions, while WGS detected 20 more events not covered by TPS. Tumor cell purity and mean ploidy showed significant correlation between WGS and TPS (Pearson r 0.524 and 0.468 respectively, p<0.05). Finally, WGS detected 12 more druggable events. From 5 samples, WGS detected some druggable events but TPS could not find any. My study demonstrates that WGS has equivalent capability for the detection of classical driver events with established importance. Moreover, WGS is superior for pattern-based features and detection of druggable driver events.