Genetic Artificial Intelligence in Gastrointestinal Disease: A Systematic Review

Area of Science:

• Bioinformatics
• Genetics
• Artificial Intelligence

Background:

• Predictive and explainable AI is increasingly applied to bioinformatics data, including single nucleotide polymorphism (SNP) information, for disease diagnosis.
• There is a need for reviews on the progress of genetic AI in the early diagnosis of gastrointestinal diseases (GID).

Purpose of the Study:

• To systematically review recent advancements in the application of genetic artificial intelligence for the early diagnosis of gastrointestinal diseases.

Main Methods:

• A systematic review of ten original studies published in PubMed from 2010 onwards.
• Studies included GID or associated diseases as the dependent variable, artificial intelligence as the intervention, and accuracy, AUC, or variable importance as outcomes.
• English language publications were considered.

Main Results:

• Performance metrics varied, with accuracy ranging from 79-100% and Area Under the Curve (AUC) from 63-98%.
• Random forest models demonstrated high performance, achieving 98% AUC for inflammatory bowel disease classification and 99% R-square for gut microbiome SNP regression.
• Key SNPs and other factors like age, BMI, and microbiome were identified as important predictors for GID.

Conclusions:

• Genetic artificial intelligence serves as an effective and non-invasive decision support system for diagnosing gastrointestinal diseases.
• The review identified significant SNPs and other variables crucial for predicting GID, although no deep learning studies were found.