Gut Microbiota Analysis and In Silico Biomarker Detection of Children with Autism Spectrum Disorder across Cohorts.

The study of human gut microbiota has attracted increasing interest in the fields of life science and healthcare. However, the complicated and interconnected associations between gut microbiota and human diseases are still difficult to determine in a predictive fashion. Artificial intelligence such as machine learning (ML) and deep learning can assist in processing and interpreting biological datasets. In this study, we aggregated data from different studies based on the species composition and relative abundance of gut microbiota in children with autism spectrum disorder (ASD) and typically developed (TD) individuals and analyzed the commonalities and differences of ASD-associated microbiota across cohorts. We established a predictive model using an ML algorithm to explore the diagnostic value of the gut microbiome for the children with ASD and identify potential biomarkers for ASD diagnosis. The results indicated that the Shenzhen cohort achieved a higher area under the receiver operating characteristic curve (AUROC) value of 0.984 with 97% accuracy, while the Moscow cohort achieved an AUROC value of 0.81 with 67% accuracy. For the combination of the two cohorts, the average prediction results had an AUROC of 0.86 and 80% accuracy. The results of our cross-cohort analysis suggested that a variety of influencing factors, such as population characteristics, geographical region, and dietary habits, should be taken into consideration in microbial transplantation or dietary therapy. Collectively, our prediction strategy based on gut microbiota can serve as an enhanced strategy for the clinical diagnosis of ASD and assist in providing a more complete method to assess the risk of the disorder.