AI-Driven Approaches to Overcoming Tumor Heterogeneity in Breast Cancer: Modelling Resistance and Therapy Outcomes

Abstract

Breast cancer is a highly heterogeneous disease consisting of distinct molecular subtypes and differential response to treatment. The major challenge associated with predicting treatment responses and overcoming resistance mechanisms in tumor heterogeneity (genetic, phenotypic, and molecular diversity of a single tumor) remains yet to be overcome. Now artificial intelligence (AI) and machine learning (ML) promise to help us address these challenges by modelling complex interactions within the tumor. Using large-scale genomic, transcriptomic, imaging and clinical data, AI can incorporate cancer heterogeneity by characterizing and identifying key drivers for cancer heterogeneity and predicting resistance mechanisms. They help to generate more accurate prediction of chemotherapy, targeted therapy, and immunotherapy treatment responses and it can pave the way for more personalized treatment strategy. But AI is also capable of monitoring tumor evolution over time, and uncovering real-time insights about how effective treatment is and if relapse is starting, early. While promising, practitioners have been slow to adopt AI models for clinical work, due to data quality, model interpretability and compatibility in the current healthcare workflow. In translating AI based tools for practice, these challenges will need to be addressed. In this paper, we discuss how AI may be used to target tumor heterogeneity, model resistance mechanisms and optimize treatment responses in breast cancer as well as presenting obstacles to clinical implementation of such methods.