Comparative analysis of tumor content estimation methods based on simulated tumor samples identified their impact on somatic variant detection in cancer whole genome sequencing

Whole genome sequencing (WGS) in cancer genomics has become widespread with recent technological innovations, and the amount and types of information obtained from WGS are increasing rapidly. Appropriate interpretation of results is becoming increasingly important in clinical applications. This study aimed to evaluate the accuracy of tumor content estimation and its impact on somatic variant detection, using 100 simulated tumor samples covering 10-100% tumor content constructed from the sequencing data of cell line models. Extensive analysis revealed that the estimation results varied among computational analytical methods. Notably, there was a large discrepancy in low tumor content (<= 30%). The reproducibility decreased in cases wherein chromosome-scale copy number changes were observed in normal cells. The minimum tumor content required to detect somatic alterations was estimated to be 10-30%. Identification of whole genome doubling was achieved with the lowest tumor content, followed by single nucleotide variation/insertion or deletion, structural variation, and copy number variation. Tumor content had a significantly higher impact on the false negatives than the false positives in variant calls. Results should be interpreted cautiously for samples wherein tumor content is a concern. These results can form the basis of developing important guidelines for evaluating cancer WGS.