Predicting Neoadjuvant Therapy Response in Breast Cancer Patients: A Multi-Omics and Machine Learning Perspective

Breast Cancer (BC) treatment response varies due to underlying heterogeneity. Personalized therapy based on multi-omics profiling enhances efficacy by identifying patientspecific biomarkers and optimizing strategies. Multiomics integrates diverse biological data to understand mechanisms and enable customized treatments. Advances in ML and DL revolutionize BC therapy response prediction, leveraging multi-omics to improve precision, identify biomarkers, and refine strategies, reducing morbidity and mortality. This study presents a comparative analysis of multi-omics-dependent models for predicting neoadjuvant therapy response, highlighting techniques like DeepSurv, Gradient Boosting Machine (GBM), and Weighted MultiSource Canonical Correlation Analysis (WMSCCA). These models use data sets such as TCGA, METABRIC, and ICGC to boost predictive power. DL enables automated feature extraction, while ML offers interpretability for balanced predictive analytics. Despite progress, challenges remain, including data limitations, lack of external validation, and interpretability issues.