Multiomics Analysis for Predicting Pathological Complete Response in Triple-Negative Breast Cancer and Reflecting Tumor Heterogeneity.

Background: Heterogeneity in triple-negative breast cancer (TNBC) leads to different responses to neoadjuvant chemotherapy (NAC). NAC-resistant TNBC is often associated with higher risk of recurrence and poor prognosis. This study developed and validated a novel radiomics-based model to predict pathological complete response (pCR) to NAC and reflect tumor heterogeneity in TNBC.

Methods: 169 TNBC patients who underwent NAC between 2013 and 2023 were screened as a training cohort. A validation cohort and 2 cohorts containing RNA-seq data were also included. Radiomics features were extracted from dynamic contrast enhanced MRI (DCE-MRI) for model construction. Based on the model, we calculated the radiomics score (Rad-score) of each patient. The predictive capacity of the model was evaluated by area under receiver operating characteristic (ROC) curves. RNA-seq data was used to evaluate drug sensitivity, enriched pathways, and tumor microenvironment (TME) characteristics.

Results: The radiomics model can predict pCR in both the training cohort (AUC = 0.902) and validation cohort (AUC = 0.775). The high Rad-score subgroup exhibited better response to chemotherapy and better prognosis. Immune activation-related pathways were also enriched in the high-score subgroup. The low-score subgroup showed enrichment of TGF-β-related pathways and was more sensitive to TGF-β inhibitor. The model can also identify immune phenotypes (AUC = 0.85). The high Rad-score subgroup had abundant immune cell infiltration, while the low Rad-score subgroup was lacking immune cells in TME.

Conclusion: The model can effectively predict the pCR of TNBC and reflect tumor heterogeneity. Chemotherapy combined with targeting the TGF-β pathway is a potential strategy to overcome drug resistance in TNBC.