Towards Intelligent Cancer Subtyping: Integrating Multi-Omics Data Using Diverse Machine Learning Strategies

Abstract

Precision oncology heavily relies on dividing cancer into different subtypes. Singleomics data do not accurately reflect the complex ways in which molecules interact and collaborate in tumor growth and development. Advances in high-throughput methods have enabled the generation of genomic, transcriptomic, epigenomic, and proteomic data, providing new opportunities to examine cancer. However since various types of data must be used together, there are large computational obstacles involved. In this paper,we look intoways that combine different machine learning (ML) methods for integrating various kinds of data to find cancer subtypes. We describe the main data forms, methods for data integration, model design strategies, testing techniques, and new difficulties. In addition, we address the impact of biological significance, easy understanding, and medical application on the future of intelligent cancer subtyping.