Evaluation of potential of targeted sequencing through mutational signature simulation

Targeted sequencing is critical in cancer diagnosis, treatment selection, and monitoring. However, the effectiveness of these methods for reflecting whole-exome sequencing (WES)-level mutational signatures remains unclear. Therefore, we addressed this issue through simulation-based analysis to clarify how well targeted sequencing can reproduce WES-level mutational signatures.

We compared the correlation and similarity of mutational signatures between whole-exome sequencing-level mutation data and downsampled data for gene sets targeted by each sequencing method in 13 cancer types. Additionally, a similarity analysis of the mutational signatures was conducted using randomly downsampled data.

The comparison between whole-exome sequencing and targeted sequencing showed a low correlation based on Pearson's correlation coefficient but a high similarity based on the Dice similarity index. As a result of the downsampling of data with cancer-related genes and whole genes to evaluate similarity, the cancer-related gene random set showed high similarity when 200-400 genes were selected. However, the whole-genome random set required 2-3 times as many genes as the cancer-related gene random set to show high similarity. Among cancer types, colorectal and lung cancers demonstrated high similarity with fewer downsampled genes, whereas breast and prostate cancers required more downsampled genes to achieve high similarity.

This study demonstrated that current clinically used targeted sequencing methods can reflect whole-exome sequencing-level mutational signatures. This suggests that considering the cancer type and average number of gene mutations in each patient when selecting targeted sequencing methods can lead to more effective treatment choices.