Integrative multi-omics and machine learning/deep learning approaches in cancer knowledge discovery: A scoping review

• Genomics, epigenomics, and transcriptomics were the most commonly integrated, identifying biologically meaningful signatures and revealing cancer subtypes of different cancer types. • Machine learning methods have been increasingly utilized for predicting survival rates, disease progression, and immune responses. • This review highlights a growing trend in applying ML and DL techniques for multi-omics integration in cancer research. The interest in AI applications in the medical field, particularly cancer research, is growing. Ma- chine learning (ML) and deep learning (DL) approaches have revolutionized data analysis in oncology. Integrating multiomics, such as transcriptomics, genomics, or proteomics offers a unique opportunity to understand the cancer complexity and identify novel therapeutic strate- gies. This scoping review aims to evaluate the effectiveness of various approaches in the cur- rent literature on multiomics integration. It focuses on describing models, methods, the applica- tion of ML/DL to discover new insights into cancer and identifying key themes and knowledge gaps to clarify the current literature across 56 included studies. These studies highlighted the methods or approaches used for predicting the prognostic or diagnostic variables, for cancer subtyping and classification tasks. The analysis showed a significant diversity in the chosen designs and data types, making it difficult to compare studies. These studies highlight the cru- cial role of multiomics integration in predicting clinical outcomes and classifying subtypes, thus advancing our understanding of cancer. This scoping review has summarized and analyzed existing approaches and methods, highlighting the need for further standardization of integra- tion methods for prediction and subtype classification.