A scoring screening system based on machine learning for immunotherapy beneficiaries in triple-negative breast cancer

Abstract Background: The role of the tumor microenvironment (TME) in predicting prognosis and therapeutic efficacy has been demonstrated. Nonetheless, no systematic studies have focused on TME patterns or their function in the effectiveness of immunotherapy in triple-negative breast cancer (TNBC).Methods: In this study, we comprehensively estimated the TME infiltration patterns of 491 TNBC patients from four independent cohorts, and three cohorts that received immunotherapy were used for validation. The TME subtypes were comprehensively evaluated based on immune cell infiltration levels in TNBC, and the TME score was identified and systematically correlated with representative tumor characteristics. Eventually, we sequenced 80 TNBC samples as an external validation cohort to make our conclusions more convincing.Results: Two distinct TME subtypes were identified and were highly correlated with immune cell infiltration levels and immune-related pathways. More representative TME scores calculated by machine learning could reflect the fundamental characteristics of TME subtypes and predict the efficacy of immunotherapy and prognosis of TNBC patients. A low TME score, characterized by activation of immunity and ferroptosis, indicated an activated TME phenotype and better OS. A high TME score, characterized by the activation of immunosuppressive cells such as CAFs, MDSCs, TAM-M2, cancer stem cells, and a lack of adequate immune infiltration, indicated an immune-suppressed TME phenotype and poorer survival. Low TME scores showed a better response to immunotherapy in TNBC by TIDE analysis and sensitivity to multiple drugs in GDSC analysis. A low TME score showed a significant therapeutic advantage in patients in the three immunotherapy cohorts.Conclusions: TME subtypes played an essential role in assessing the diversity and complexity of the TME in TNBC. The TME score could be used to evaluate the TME of an individual tumor to enhance our understanding of the TME and guide more effective immunotherapy strategies. Depicting a sweeping landscape of the TME characteristics of TNBC may therefore help to interpret the responses of TNBC to immunotherapies and provide new strategies for the treatment of cancers.