Abstract 389: Feasibility of normalizing TMB for tumor tissue samples with low purity

Abstract

Abstract Background: Tumor mutation burden (TMB) is currently recognized as one of the major indicators for the efficacy of immunotherapy. In the meantime, tumor purity as one of the confounders can affect unbiased estimation of TMB. The aim of this study is to profile the correlation between tumor purity and TMB using clinically sequenced tumor samples, and explore the feasibility of restoring TMB for objective therapeutic guidance. Method: We collected 68 lung cancer tissue samples with purity ranging from40% to 70% assessed by IHC. These tumor samples alone with their matched WBC were sequenced using customized panel covering 831 cancer genes. Raw data (FASTQ file) of tumor samples were then diluted into four tiers (40%, 30%, 20% and 15%) using FASTQ files of their matched WBC samples. Standard analytical pipeline was applied to all original as well as diluted data for detection of point somatic mutations. TMB was calculated after proper mutation filtration. Result: The TMB of 68 lung cancer samples ranges from 1.5 to 42.3. TMB remains unchanged for most samples when purity was diluted to 40%, indicating that tumor with 40% purity can be used for unbiased TMB estimation. TMB drops 14%, 35% and 44% on average when sample purity was diluted to 30%, 20% and 15%. We hypothetically set the third quartile of TMB (TMB = 6) as cutoff of immunotherapy effective. when tumor purity was diluted to 30%, 20% and 15%, only 89%, 75% and 61% samples remain immunotherapy effective respectively. To restore objective TMB, we applied a simple method. For each sample, we calculated ratio of TMB between diluted purity and purity of 40%. For each diluted purity tier, the mean TMB ratio, standard deviation and confidence interval were calculated among 68 tumor samples. The lower bound was used as ‘purity specific coefficient' to restore TMB. Using TMB of 6 as cutoff, this approach resulted in 6% (2) over-estimation and 12% (4) correction for samples of 30% purity, 3% (1) over-estimation and 22% (7) correction for samples of 20% purity, 15% (5) over-estimation and 34% (11) correction for samples of 15% purity. For each diluted purity tier, this method corrects more samples eligible for immunotherapy than false positive cases. Conclusion: We applied an in silico approach to explore the correlation between TMB and purity of tumor samples. Results showed that TMB was under-estimated for samples with purity lower than 40%. The degree of under-estimation is generally proportional to the decrease of tumor purity. A simple method can learn a purity specific coefficient from a set of tumor samples and improve TMB estimation for low purity tumor samples. Citation Format: Cheng Yan, Tao Fu, Junhui Yang, Qin Zhang, Yufei Yang, Changshi Du. Feasibility of normalizing TMB for tumor tissue samples with low purity [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 389.