171MO - Unraveling iCAF and tumor cell communication: Advancing novel adjuvant immunotherapy response and prognostic prediction via single-cell analysis and multimodal data-based models

Background

Accurate prediction of neoadjuvant immunotherapy efficacy is crucial for improving outcomes in triple-negative breast cancer (TNBC) patients. This study aims to identify key cancer cell and cancer-associated fibroblast (CAF) subsets closely linked to resistance to neoadjuvant immunotherapy and to develop a multimodal prognostic prediction model.

Methods

In this study, tumor samples from patients diagnosed with TNBC who received neoadjuvant immunotherapy were analyzed using single-cell RNA sequencing. To dissect the TNBC ecosystem in patients with varying responses to neoadjuvant immunotherapy, we conducted a systematic analysis of cellular compositions across individual TNBC samples. A prognostic model based on cell interactions was developed using deep learning algorithm. The model utilized 139 genes from 171 ligand-receptor pairs that were significant in interactions between inflammatory CAFs (iCAFs) and Cancer_cell_7. A multimodal data-based model was employed to discriminate prognostic-related Cancer_cell-iCAF interactions (PCAF). The model incorporated features related to DNA methylation, somatic mutations, miRNA, immune cell infiltration, and whole slide images.

Results

We identified specific cancer cell and CAF subsets strongly associated with neoadjuvant immunotherapy resistance and poor clinical outcomes. The multimodal data-based prognostic model, DeepClinMed-PCAF, served as a robust risk score for predicting poor prognosis and demonstrated high accuracy in breast cancer patients. In a training cohort of 460 patients, 98 individuals with high PCAF scores exhibited significantly poorer overall survival (HR 9.04; P < 0.001; AUC 0.81 at 60 months). The model also showed strong performance in a validation cohort (AUC 0.79 at 60 months) and an external test cohort (AUC 0.88 at 60 months).

Conclusions

This study underscores the potential of PCAF in identifying patients suitable for immunotherapy. The comprehensive understanding of cell interactions and the prognostic model presented here provide valuable insights for personalized treatment approaches in advanced TNBC patients undergoing neoadjuvant immunotherapy.

Legal entity responsible for the study

The authors.

Funding

National Natural Science Foundation of China.

Disclosure

All authors have declared no conflicts of interest.