Genetic Artificial Intelligence in Gastrointestinal Disease: A Systematic Review

Background/Objectives: The application of predictive and explainable artificial intelligence to bioinformatics data such as single nucleotide polymorphism (SNP) information is attracting rising attention in the diagnosis of various diseases. However, there are few reviews available on the recent progress of genetic artificial intelligence for the early diagnosis of gastrointestinal disease (GID). The purpose of this study is to complete a systematic review on the recent progress of genetic artificial intelligence in GID. Methods: The source of data was ten original studies from PubMed. The ten original studies were eligible according to the following criteria: (participants) the dependent variable of GID or associated disease; (interventions/comparisons) artificial intelligence; (outcomes) accuracy, the area under the curve (AUC), and/or variable importance; a publication year of 2010 or later; and the publication language of English. Results: The performance outcomes reported varied within 79-100 for accuracy (%) and 63-98 for the AUC (%). Random forest was the best approach (AUC 98%) for the classification of inflammatory bowel disease with 13 single nucleotide polymorphisms (SNPs). Similarly, random forest was the best method (R-square 99%) for the regression of the gut microbiome SNP saturation number. The following SNPs were discovered to be major variables for the prediction of GID or associated disease: rs2295778, rs13337626, rs2296188, rs2114039 (esophageal adenocarcinoma); rs28785174, rs60532570, rs13056955, rs7660164 (Crohn's disease early intestinal resection); rs4945943 (Crohn's disease); rs316115020, rs316420452 (calcium metabolism); rs738409_G, rs2642438_A, rs58542926_T, rs72613567_TA (steatotic liver disease); rs148710154, rs75146099 (esophageal squamous cell carcinoma). The following demographic and health-related variables were found to be important predictors of GID or associated disease besides SNPs: age, body mass index, disease behavior, immune cell type, intestinal microbiome, MARCKS protein, smoking, and SNP density/number. No deep learning study was found even though deep learning was used as a search term together with machine learning. Conclusions: Genetic artificial intelligence is effective and non-invasive as a decision support system for GID.