GAN-GMHI: Generative Adversarial Network for high discrimination power in microbiome-based disease prediction

Abstract

Gut microbiome-based health index (GMHI) has been applied with success, while the discrimination powers of GMHI varied for different diseases, limiting its utility on a broad-spectrum of diseases. In this work, a generative adversarial network (GAN) model is proposed to improve the discrimination power of GMHI. Built based on the batch corrected data through GAN, GAN-GMHI has largely reduced the batch effects, and profoundly improved the performance for distinguishing healthy individuals and different diseases. GAN-GMHI has provided a solution to unravel the strong association of gut microbiome and diseases, and indicated a more accurate venue toward microbiome-based disease monitoring. The code for GAN-GMHI is available at https://github.com/HUST-NingKang-Lab/GAN-GMHI. ImportanceThe association of gut microbiome and diseases has been proven for many diseases, while the transformation of such association to a robust and universal disease prediction model has remained illusive, largely due to the batch effects presents in multiple microbiome cohorts. Our analyses have indicated a plausible venue, which is based on GAN technique, towards batch effect removal for microbiome datasets. GAN-GMHI is a novel method built based on the batch corrected data through GAN, as well as GMHI for prediction of a broad-spectrum of diseases.