Recent advancements in artificial intelligence - driven breast cancer molecular subtypes classification using multi-omics: A comprehensive review

Breast cancer is one of the heterogeneous diseases comprising various molecular subtypes. All molecular subtypes have different characteristics and behave differently to treatment response, prognosis and therapy. Accurate and precise classification of breast cancer molecular subtypes is crucial to know how breast cancer behaves, grows and responds to treatment and prognosis on the molecular level. There are a few existing studies conducted on breast cancer molecular subtypes classification, either using mono-omics or multi-omics, while lacking systematic comparisons of both. However, there is a need to know the performance and differences of mono-omics and multi-omics high-throughput technologies for breast cancer molecular subtypes classification, including the taxonomy, heterogeneity, causes, risk factors and unique molecular characterization. Therefore, to overcome these issues, this comprehensive review provides a structured synthesis of the current state of research on breast cancer molecular subtypes classification. Artificial Intelligence (AI)-driven Machine Learning (ML) and Deep Learning (DL) models, are employed for breast cancer molecular subtypes classification, mainly focusing on mono-omics and multi-omics. The analysis of this review shows that multi-omics technologies have great potential for the accurate and precise classification of breast cancer compared to mono-omics. It not only provides a detailed structure of the breast cancer molecular subtypes but also provides a comprehensive view of tumor progression, growth dynamics, aggressiveness, and underlying biological mechanisms. The correct integration of multi-omics data types and variants plays a significant role in classifying breast cancer molecular subtypes. Based on the extensive analysis of the existing studies, some of the main challenges that still exists remain in the classification of breast cancer molecular subtypes, include high dimensionality of multi-omics data, overfitting, data imbalance, models overperformance on minority classes, high correlation and overlapping, computational complexity, accurate integration of multi-omics data types and variants, analysis of the misclassification patterns and accurate classification of breast cancer molecular subtypes.