Development of a tertiary lymphoid structure-based prognostic model for breast cancer: integrating single-cell sequencing and machine learning to enhance patient outcomes

Abstract

Background: Breast cancer, a highly prevalent global cancer, poses significant challenges, especially in advanced stages. Prognostic models are crucial to enhance patient outcomes. Tertiary lymphoid structures (TLS) within the tumor microenvironment have been associated with better prognostic outcomes.

Methods: We analyzed data from 13 independent breast cancer cohorts, totaling over 9,551 patients. Using single-cell RNA sequencing and machine learning algorithms, we identified critical TLS-associated genes and developed a TLS-based predictive model. This model stratified patients into high and low-risk groups. Genomic alterations, immune infiltration, and cellular interactions within the tumor microenvironment were assessed.

Results: The TLS-based model demonstrated superior accuracy compared to traditional models, predicting overall survival. High TLS patients had higher tumor mutation burden and more chromosomal alterations, correlating with poorer prognosis. High-risk patients exhibited a significant depletion of CD4⁺ T cells, CD8⁺ T cells, and B cells, as evidenced by single-cell and bulk transcriptomic analyses. In contrast, immune checkpoint inhibitors demonstrated greater efficacy in low-risk patients, whereas chemotherapy proved more effective for high-risk individuals.

Conclusions: The TLS-based prognostic model is a robust tool for predicting breast cancer outcomes, highlighting the tumor microenvironment's role in cancer progression. It enhances our understanding of breast cancer biology and supports personalized therapeutic strategies.

Keywords: breast cancer; immune microenvironment; machine learning algorithms; prognostic prediction models; tertiary lymphoid structures.