Multi-cohort and single-cell profiling of aging genes reveals prognostic and therapeutic targets in breast cancer

TL;DR

A machine learning model based on aging-related genes helps predict breast cancer outcomes and treatment responses.

Contribution

A novel aging gene signature (MLAG) is developed that outperforms existing models in predicting breast cancer prognosis.

Findings

The MLAG score consistently stratifies breast cancer patients by survival risk across multiple datasets.

High MLAG tumors exhibit genomic instability and immunosuppressive features, while low MLAG tumors show enhanced immune infiltration.

MLAG predicts immunotherapy response and identifies potential therapeutic strategies like panobinostat.

Abstract

Aging-related transcriptional programs shape breast cancer progression, immune regulation, and therapeutic response. We integrated curated aging-associated genes with 108 survival modeling strategies to derive a machine learning-based aging gene signature. Trained in the TCGA-BRCA cohort and evaluated across twelve independent datasets, the eight-gene MLAG score consistently stratified patients by survival risk across clinical contexts. High MLAG exhibited increased genomic instability, including elevated tumor mutational burden and copy number alterations, together with distinct regulatory and intercellular communication patterns. In contrast, low MLAG showed enhanced immune infiltration, coordinated microenvironment signaling, and higher immune checkpoint expression. Integration of bulk and single-cell transcriptomic analyses localized MLAG-associated aging programs primarily to malignant epithelial cells and linked them to aneuploidy and stress-response pathways. The MLAG score associated with immune checkpoint blockade response and supported identification of candidate therapeutic strategies. These findings highlight the relevance of aging-associated transcriptional states in breast cancer prognosis and treatment stratification.

•A machine learning aging gene signature robustly stratifies breast cancer prognosis•The MLAG model outperforms 100 published signatures across multiple cohorts•High MLAG tumors show genomic instability and immunosuppressive microenvironments•MLAG predicts immunotherapy response and identifies panobinostat as a candidate therapy

A machine learning aging gene signature robustly stratifies breast cancer prognosis

The MLAG model outperforms 100 published signatures across multiple cohorts

High MLAG tumors show genomic instability and immunosuppressive microenvironments

MLAG predicts immunotherapy response and identifies panobinostat as a candidate therapy

Computational bioinformatics; Cancer; Omics

Figures