Interpretable Machine Learning Algorithms Reveals Gut Microbiome Features Associated with Atopic Dermatitis

Jingtai Ma ^1,2* yiting fang ^1,2* shiqi li ^1,2 Lilian Zeng ² Siyi Chen ² Zhifeng Li ² Guiyuan Ji ² Xing-Fen Yang ¹ wei wu ^1,3

• ¹ NMPA Key Laboratory for Safety Evaluation of Cosmetics, Guangdong Provincial Key Laboratory of Tropical Disease Research, School of Public Health, Southern Medical University, Guangzhou, China
• ² Guangdong Provincial Institute of Public Health, Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, Guangdong Province, China
• ³ Guangdong Provincial Center for Disease Control and Prevention, Guangzhou, China

The "gut-skin axis" has been proposed to play an important role in the development and symptoms of atopic dermatitis. So, we construct an interpretable machine learning framework to screen key gut flora quantificationally.The 16s dataset with centered log-ratio transformation was explored using five different machine learning models, including random forest, light gradient boosting machine, extreme gradient boosting, support vector machine with radial kernel and logistic regression. Interpretable methods based on machine learning, such as SHAP value, were used to estimate significant features associated with atopic dermatitis.Results: Random forest performed better than the other "tree" models in the validation partitions. The SHAP global dependency plot screened that Bifidobacterium ranked as the strongest predictive factor for all prediction horizons, though the SHAP values of some features were still higher in support vector machine and logistic regression. The SHAP partial dependency plot for "tree" models indicated that best segmentation point for the Bifidobacterium was further away from the origin than other features in respective models, quantitatively reflecting the difference of gut microbiota.The machine learning models combined with SHAP could be used to screen key gut flora quantificationally in atopic dermatitis patients and give doctors an intuitive knowledge of 16s rRNA sequencing data to support precision medicine in care and recovery.