Abstract 4283: scMSI: Accurately detect the sub-clonal micro-satellite instability by an integrative Bayesian model

Abstract Microsatellite instability (MSI) is an important genomic biomarker for cancer diagnosis and treatment. For some cancers with high-degree of heterogeneity, e.g. endometrial cancer, the existing approaches always fail to identify the micro-satellite instability on one or multiple sub-clones, which would deprive the chance for patients to benefit from treatments. However, it is a computational challenge to estimate the sub-clonal MSI because multiple sub-clones may share the genomic status. Herein, in this paper, we propose an accurate and efficient algorithm, named scMSI, to estimate the sub-clonal microsatellite status. scMSI adopts an alternating iterative model to de-convolute the length distribution, which is a mixture of sub-clones. During the deconvolution, an optimized division of each sub-clone is achieved by a heuristic algorithm, which is bounded to the clonal proportions best consistent with the known clonal structure. To evaluate the performance, we conducted a series of experiments on simulation datasets. The results supported that scMSI solved the detection problem of MSI on sub-clones. It outperforms the existing approaches on multiple metrics. In addition, we collected a cohort of 16 endometrial cancer patients, who have positive responses on the treatment but with negative MSI status. We sequenced these patients. scMSI reported MSI on sub-clones according the sequencing data, which are further validated by the conclusions on immunohistochemistry. Thus, scMSI could provide a powerful tool for MSI analysis. Citation Format: Yuqian Liu, Yan Chen, Huanwen Wu, Xuanping Zhang, Xin Lai, Xin Yi, Zhiyong Liang, Jiayin Wang. scMSI: Accurately detect the sub-clonal micro-satellite instability by an integrative Bayesian model. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4283.