Advancing Cancer Biomarker Discovery Through Multi-Omics, Bioinformatics, and Artificial Intelligence: A Systematic...

The rapid accumulation of multi-omics data provides increasing opportunities for cancer biomarker discovery, but the heterogeneity and high dimensionality of these datasets remain major challenges for data integration and analysis. This review examines the application of bioinformatics and artificial intelligence for multi-omics integration in cancer biomarker research. Common multi-omics data paradigms and integration strategies are discussed, followed by a review of the bioinformatics, machine learning and deep learning approaches applied across 84 included studies. Cross-study comparison showed limited recurrence among the reported biomarkers, with only 15 biomarkers identified in at least four studies. Further assessment of the 13 recurrent protein-coding biomarkers showed significant protein-protein interaction enrichment and functional convergence in cancer-related processes, particularly PI3K/AKT and ERBB2-related signaling. This review also highlights current challenges, including missing data, heterogeneity across datasets, limited independent validation and the interpretability of artificial intelligence models. Future research may benefit from integrating multi-omics data with clinical information, pathological imaging, electronic health records, single-cell multi-omics and radiomics, together with broader datasets and greater use of independent validation to support the translation of multi-omics biomarker findings into clinical applications.