Development of diagnostic and prognostic molecular biomarkers in hepatocellular carcinoma using machine learning: A systematic review

Abstract

Hepatocellular carcinoma (HCC) is a leading cause of cancer-related mortality and morbidity worldwide. Machine learning (ML) tools have been developed in recent years to generate diagnostic and prognostic molecular biomarkers for this high-fatality cancer. To delineate the landscape of ML in HCC, we performed a systematic search of Ovid Medline, Ovid Embase, Cochrane Database of Systematic Reviews (Ovid) and Cochrane CENTRAL (Ovid) to identify studies of HCC molecular biomarkers using ML strategies. In total, 75 studies met our inclusion criteria, 53 of which were pertinent to diagnosis of HCC and 22 of which were pertinent to prognostication of HCC. Genomic, transcriptomic, epigenomic, proteomic and metabolomic signatures were derived using various ML techniques (supervised, unsupervised and deep learning approaches) using serum, urine and tissue samples of HCC. The ML algorithms achieved a sensitivity of up to 95% for the diagnosis of HCC. Through pathway analysis of the signatures derived by ML tools, we identified regulators of epithelial-mesenchymal transition and the cancer pathway Ras/Raf/MAPK as being particularly prognostic of HCC outcome. The application of ML to molecular data in HCC has thus far resulted in the generation of highly sensitive diagnostic and prognostic signatures. In future, development of ML algorithms that incorporate clinical, laboratory, alongside molecular features will be needed to fulfil the promise of personalized HCC diagnosis and treatment.