Machine Learning for Diagnosis of Diseases with Complete Gene Expression Profile

This paper considers the use of machine learning for diagnosis of diseases that is based on the analysis of a complete gene expression profile. This distinguishes our study from other approaches that require a preliminary step of finding a limited number of relevant genes (tens or hundreds of genes). We conducted experiments with complete genetic expression profiles (20 531 genes) that we obtained after processing transcriptomes of 801 patients with known oncologic diagnoses (oncology of the lung, kidneys, breast, prostate, and colon). Using the indextron (instant learning index system) for a new purpose, i.e., for complete expression profile processing, provided diagnostic accuracy that is 99.75% in agreement with the results of histological verification.

Automated summary

This paper explores the use of machine learning for disease diagnosis based on analysis of complete gene expression profiles. Unlike other approaches, which require a preliminary step of identifying a limited number of relevant genes, this method achieves high diagnostic accuracy without such limitations.