人工智能驱动的新型个性化肿瘤疫苗设计及其作用机制研究

肿瘤特异性抗原是人体免疫系统抑制肿瘤发生、发展的基础。源于肿瘤突变蛋白的新生抗原可绕过中枢免疫耐受机制成为肿瘤特异性抗原。利用肿瘤体细胞突变信息和新生抗原预测的人工智能算法，可以发现个体化的新生抗原，并可进一步设计成为肿瘤个体化疫苗。申请人课题组与国际同步开展了人工智能驱动个性化肿瘤疫苗的设计研究，并开展了约50例临床试验，虽然初步证明了疫苗的可行性、安全性和温和的有效性，但离临床的需求尚有距离。主要问题是新生抗原的预测准确性不够和特异性T细胞的激活程度不足。本项目拟利用质谱技术建立中国人群高频HLA递呈多肽的数据库，并以此扩充深度学习训练数据集，提高新生抗原预测算法的准确性；基于我们近期的发现（Cell, 2019）设计新型个性化肿瘤疫苗及相应的人工APC，引入T细胞第二信号通路，进一步提高T细胞的激活和扩增效率，使之兼具T细胞刺激特异性和有效性，实现对肿瘤细胞的高效特异性杀伤。

Tumor-specific antigens (TSAs) are the basis for the human immune system to suppress tumorigenesis and development. Neoantigens derived from tumor mutant proteins can bypass the central immune tolerance mechanism and become TSAs. Taking advantage of the artificial intelligence algorithm of tumor somatic mutation information and neoantigen prediction, individual neoantigens can be discovered and further designed into personalized tumor vaccines. The applicant's research group has carried out the research of personalized tumor vaccines driven by artificial intelligence in parallel with the international, and carried out about 50 clinical trials. Although the feasibility, safety and mild effectiveness of the vaccine have been initially proved, it is still far from the clinical needs. The main problems are insufficient prediction accuracy of neoantigens and insufficient activation of specific T cells. This project intends to utilize mass spectrometry technology to establish a database of high-frequency HLA presentation peptides from Chinese, and to expand the deep learning training data set to improve the accuracy of the neoantigen prediction algorithm. Based on our recent findings (Cell, 2019), we designed a novel personalized tumor vaccine and corresponding artificial APC that introduced the second signaling pathway of T cells to further enhance the activation and expansion efficiency of T cells, which make them both specific and effective for T cell stimulation, thereby achieving efficient and specific killing of tumor cells.