Explainable Artificial Intelligence (XAI) for Microbiome Data Analysis in Autistic Spectrum Disorder

Motivation Autism Spectrum Disorders (ASD) are heterogeneous neurodevelopmental conditions characterized by deficits in communication and social interaction, as well as repetitive behaviours and interests. The microbiota plays an important role in regulating normal host physiology, metabolism, nutrition, and brain function. A growing number of researches have revealed the ability of the intestinal microbiota to induce signals through the ”microbiota-intestine-brain axis”. Machine learning is a subset of artificial intelligence (AI) that focuses on enabling machines to learn from data and improve their performance on specific tasks without being explicitly programmed. This type of statistical framework lends itself to multivariate analyses such as the study of the mycobiome in humans. Explainable Artificial Intelligence (XAI) is a field that seeks to develop machine learning models that can provide clear, interpretable explanations for their decisions. Recently, there has been a growing interest in applying XAI methods to the study of the microbiome, the complex community of microorganisms that inhabit the human body. Methods In particular, we used XAI to analyze the microbiome of children with Autistic Spectrum Disorder (ASD), a condition that is known to be associated with alterations in the gut microbiome. For these purposes, we used a cohort of 254 children aged 1-13 years, in particular 143 affected by Autism Spectrum Disordered (ASD) and 111 control subjects (TD). The framework implemented is composed of two parts: (i) the classification using XGBoost classifiers;(ii) the identification of the most important features in the classification of each single patient by using the SHAP algorithm. Results The XGBoost achieved the following performance: accuracy (0.913 ± 0.052), sensitivity (0.942 ± 0.064), specificity (0.875 ± 0.093) and precision (0.911 ± 0.063). From these results, we calculated the average ROC curve, obtaining an AUC of 0.094 ± 0.038. By extracting interpretable features and relationships from microbiome data, XAI approaches helped to identify underlying biological mechanisms that contribute to ASD pathology. Further research in this area has the potential to reveal new targets for therapies that can modulate the microbiome and improve outcomes for children with ASD. The results of the Shap analysis highlight the fact that two Phyla are most represented.