综合分析识别并验证肿瘤微环境亚型以预测肝细胞癌抗肿瘤治疗疗效

The objective of this study was to explore and verify the subtypes in hepatocellular carcinoma based on the immune (lymphocyte and myeloid cells), stem, and stromal cells in the tumor microenvironment and analyze the biological characteristics and potential relevance of each cluster.

We used the xCell algorithm to calculate cell scores and got subtypes by k-means clustering. In the external validation sets, we verified the conclusion stability by a neural network model. Simultaneously, we speculated the inner connection between clusters by pseudotime trajectory analysis and confirmed it by pathway enrichment, TMB, CNV, etc., analysis. RESULT: According to the results of the consensus cluster, we chose k = 4 as the optimal value and got four different subtypes (C1, C2, C3, and C4) with different biological characteristics based on infiltrating levels of 48 cells in TME. In univariable Cox regression, the hazard ratio (HR) value of C3 versus C1 was 2.881 (95% CI: 1.572-5.279); in multivariable Cox regression, we corrected the age and TNM stage, and the HR value of C3 versus C1 was 2.510 (95% CI: 1.339-4.706). C1 and C2 belonged to the immune-active type, C3 and C4 related to the immune-insensitive type and the potential conversion relationships between clusters. We established a neural network model, and the area under the curves of the neural network model was 0.949 in the testing cohort; the same survival results were also observed in the external validation set. We compared the differences in cell infiltration, immune function, pathway enrichment, TMB, and CNV of four clusters and speculated that C1 and C2 were more likely to benefit from immunotherapy and C3 may benefit from FGF inhibitors. DISCUSSION: Our analysis provides a new approach for the identification of four tumor microenvironment clusters in patients with liver cancer and identifies the biological differences and predicts the immunotherapy efficacy between the four subtypes.