Ensemble Machine Learning on Bulk RNA-Seq Identifies 17-Gene Signature Predicting Neoadjuvant Chemotherapy Response...

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Accurate prediction of neoadjuvant chemotherapy (NAC) response in breast cancer is crucial for treatment optimization.
• Current predictive biomarkers are insufficient, necessitating novel approaches.

Purpose of the Study:

• To develop and validate a gene expression signature for predicting pathological complete response (pCR) versus residual disease (RD) after NAC in breast cancer.
• To leverage ensemble machine learning for robust biomarker discovery.

Main Methods:

• Utilized bulk RNA-sequencing data from two cohorts (GSE163882 and GSE240671).
• Employed differential expression analysis, ensemble feature selection across five machine learning classifiers, and stacked ensemble modeling.
• Performed 10-fold repeated cross-validation and external validation with ComBat batch correction and isotonic calibration.

Main Results:

• Identified a 17-gene signature with high predictive performance (AUC 0.97 in internal testing, AUC 0.78 and PR AUC 0.85 in external validation).
• The signature demonstrated balanced accuracy of 0.71 and 0.86 sensitivity for pCR detection.
• Pathway enrichment analysis linked the signature to cell cycle regulation, DNA repair, and transcriptional control, highlighting biologically relevant genes.

Conclusions:

• The externally validated 17-gene signature serves as a biologically grounded tool for predicting NAC response in breast cancer.
• This signature has the potential to enhance precision oncology by guiding treatment decisions.