Designing gene panels for tumor mutational burden estimation: the need to shift from 'correlation' to 'accuracy'.

Wu HX1, Wang ZX1, Zhao Q2, Wang F3, Xu RH4.

Author information

1: Department of Medical Oncology, State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Sun Yat-Sen University Cancer Center, Guangzhou, 510060, People's Republic of China.
2: Bioinformatics Platform, Department of Experimental Research, State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Sun Yat-Sen University Cancer Center, Guangzhou, 510060, People's Republic of China.
3: Department of Medical Oncology, State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Sun Yat-Sen University Cancer Center, Guangzhou, 510060, People's Republic of China. wangfeng@sysucc.org.cn.
4: Department of Medical Oncology, State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Sun Yat-Sen University Cancer Center, Guangzhou, 510060, People's Republic of China. xurh@sysucc.org.cn.

Abstract

Tumor mutational burden (TMB) assessment is at the forefront in precision medicine. The TMB could represent a biomarker for immune checkpoint inhibitors (ICIs) responses. Whole exome sequencing (WES) is the gold standard to derive the TMB; while targeted next-generation sequencing panels might be more feasible. However, mainstream panels use 'correlation' (R2) between panel- and WES-based TMB to validate TMB estimation, which could be vulnerable to be distorted by cases with relatively ultra-high TMB within each cancer type. The FDA-approved FoundationOne CDx (F1CDx) panel-based TMB estimation seemed reliable (R2 ≥ 0.75) in 24 out of 33 cancer types from the Cancer Genome Atlas, but most of them were overestimated by correlation as only seven cancer types had satisfactory accuracy (the proportion of cases correctly identified as TMB-high or TMB-low using panel-based TMB) above 90%. After removing cases with relatively ultra-high TMB within each cancer type, the correlation (R2) in 16 of these 24 cancer types declined dramatically (Δ > 0.25) while all of their accuracy remained generally constant, indicating that accuracy is more robust than correlation. Similar results were also observed in other four panels. Further incorporating accuracy in panel design revealed that the minimal number of genes needed to achieve ≥ 90% accuracy varied among cancer types and correlated negatively with their TMB levels (p = 0.001). In summary, currently available panels can accurately assess TMB only in several particular cancer types; and accuracy outperformed correlation in assessing the performance of panel-based TMB estimation. Accuracy and cancer type individualization should be incorporated in designing panels for TMB estimation.