鉴定能区分癌细胞与正常周围组织细胞的最佳靶抗原仍是切费融合抗原受体（CAR）细胞疗法用于具有肿瘤内异质性的肿瘤的关键挑战。在本研究中，我们通过构建一个单细胞表达图谱，将组织复杂性解析到单个细胞的水平上，该图谱整合了来自412个肿瘤和12个正常器官的约1.4百万个肿瘤、渗透肿瘤的正常细胞和参考正常细胞。我们使用随机森林和卷积神经网络的两步筛选方法，选择最能区分恶性和正常单个细胞的基因对。基于成对基因在单个单个细胞上的组合表达模式，评估了与、或、非逻辑门所需的肿瘤覆盖率和特异性。单细胞转录组耦合表位分析验证了在卵巢癌和结直肠癌中鉴定的与、或和非开关目标。© 2023年。作者（们）独家许可Springer Nature America，Inc.。

Identification of optimal target antigens that distinguish cancer cells from normal surrounding tissue cells remains a key challenge in chimeric antigen receptor (CAR) cell therapy for tumors with intratumoral heterogeneity. In this study, we dissected tissue complexity to the level of individual cells through the construction of a single-cell expression atlas that integrates ~1.4 million tumor, tumor-infiltrating normal and reference normal cells from 412 tumors and 12 normal organs. We used a two-step screening method using random forest and convolutional neural networks to select gene pairs that contribute most to discrimination between individual malignant and normal cells. Tumor coverage and specificity are evaluated for the AND, OR and NOT logic gates based on the combinatorial expression pattern of the pairing genes across individual single cells. Single-cell transcriptome-coupled epitope profiling validates the AND, OR and NOT switch targets identified in ovarian cancer and colorectal cancer.© 2023. The Author(s), under exclusive licence to Springer Nature America, Inc.