MIRS: An AI scoring system for predicting the prognosis and therapy of breast cancer

Tumor-infiltrating immune cells (TIICs) and metastasis are crucial characteristics for tumorigenesis. However, the potential role of their combination in breast cancer (BRCA) remains elusive. Herein, on the basis of quantifying TIICs and tumor metastasis together, we established a precise prognostic scoring system named metastatic and immunogenomic risk score (MIRS) using a neural network model. MIRS showed better performance when compared with other published signatures. MIRS stratifies patients into a high risk subtype (MIRS^high) and a low risk subtype (MIRS^low). The MIRS^high patients exhibit significantly lower survival rate compared with MIRS^low patients ($$), higher response to chemotherapy, but lower response to immunotherapy. Conversely, higher infiltration level of TIICs and significantly prolonged survival ($$) are observed in MIRS^low patients, indicating sensitive response in immunotherapy. This work presents a promising indicator to guide treatment options of the BRCA population and provides a predicted webtool that is almost universally applicable to BRCA patients.