亚群特异性驱动基因对揭示异质性癌症发病机理及发展癌症精准诊疗极为关键。虽然现有研究在多样本的共有驱动基因预测方面取得了较多进展，但对亚群特异性驱动基因的预测，目前仍局限于亚群类别已知的理想情况。在亚群类别未知时，现有研究无法识别其中的关联性驱动基因，即受相互作用基因影响而异常的驱动基因。此外，对于多亚群共存驱动基因，现有研究尚不具备预测能力。为此，针对亚群类别未知问题，本项目拟探索面向肿瘤异质性数据的亚群特异性指引表征策略，通过向量化表征推断出驱动基因的亚群从属关系。在此基础上，针对关联性驱动基因的预测，发展表征向量点关联性模拟机制，进而强化对关联性驱动基因的预测能力。针对多亚群共存驱动基因的识别，构建表征空间多峰分布的离群点判定方法，实现单亚群特异性/多亚群共存驱动基因的同步化预测。本项目将为亚群特异性驱动基因分析提供新思路，为癌症临床的精准化诊断治疗提供支持。

Subgroup-specific driver genes are critical for revealing the pathogenesis of heterogeneous cancers and for developing the precision diagnosis and treatment of cancers. Although current studies have made considerable progress in the prediction of common driver genes from multiple samples, the prediction of subgroup-specific driver genes is still limited in the ideal situation that subgroup categories are already known. When the subgroup category is unknown, the current studies cannot identify the interacting driver genes, i.e., the driver genes which are abnormally affected by its interacting genes. Furthermore, for multi-subgroup coexisting driver genes, the current studies do not yet have the predictive ability. Therefore, to address the issue that the subgroup categories are unknown, this project aims to explore the subgroup-specific inducing representation strategy for tumor heterogeneous data, and build the vectorization representation to infer the subgroup affiliation of driver genes. Based on the representation results above, to predict the interacting driver genes, we develop the mechanism of the correlation simulation for the vectorization representation points, and then enhance the predictive ability of the interacting driver genes. For the recognition of multi-subgroup coexisting driver genes, we establish an outlier detection method for the multimodal distribution in the representation space, and achieve the synchronous prediction of both single-subgroup specific/multi-subgroup coexisting driver genes. This project will provide new ideas for subgroup-specific driver gene analysis, and support the precision diagnosis and treatment in cancer clinical trials.