Improving Outcomes in Hepatocellular Carcinoma through Integration of Machine Learning: Development of a Tumor-Associated Macrophage Signature.

Abstract

Introduction: Hepatocellular carcinoma (HCC) is one of the most common malignant tumors globally. Macrophages, as essential components of the immune system, play crucial roles in immune regulation, inflammation modulation, and antitumor activity. However, it remains unclear whether tumor-associated macrophages can serve as prognostic markers for HCC.

Methods: First, we identified tumor-associated macrophages based on single-cell data from GSE140228. Then, using a machine learning approach with a combination of 101 module genes, we constructed an optimal prognostic model. Subsequently, we compared our constructed model with other published prognostic models for HCC. Finally, we utilized the generated model score to predict the response to chemotherapy and immune therapy.

Results: First, we identified clusters of tumor-associated macrophages using single-cell data. Subsequently, we calculated the tumor-associated macrophage score based on module genes from the previous step. Compared to traditional clinical indicators, tumor-associated macrophage signature (TAMS) exhibits significant advantages. The TAMS C-index not only predicts overall survival, but also recurrence-free survival in HCC patients. Additionally, there was a higher prevalence of TP53 mutations in HCC patients with high TAMS. Furthermore, patients with low TAMS showed greater sensitivity to immunotherapy compared to those with high TAMS. Notably, the number and intensity of interactions between TAM and other T lymphocytes were significantly higher than those involving other cell populations. Interestingly, the high TAMS group exhibited significantly elevated levels of immune checkpoint markers and M2 macrophage markers.

Conclusion: TAMS can serve as a novel and potent tool, offering improved treatment options and prognostic assessment for patients with HCC.