Enhanced Metagenomic Deep Learning for Disease Prediction and Reproducible Signature Identification by Restructured Microbiome 2D-Representations

Metagenomic analysis has been explored for disease diagnosis and biomarker discovery. Low-sample sizes, high-dimensionality, and sparsity of metagenomic data challenge metagenomic investigations. Here, an unsupervised Microbial Embedding, Grouping, and Mapping Algorithm (MEGMA) was developed to transform metagenomic data into individualised multichannel microbiome 2D-representation by manifold learning and clustering of microbial profiles (e.g., composition, abundance, hierarchy, and taxonomy). These 2D-representations enable enhanced disease prediction by established deep ConvNet architectures, significantly outperforming the state-of-the-art machine learning and deep learning models in metagenomic benchmark datasets. These 2D-representations combined with ConvNet explainable module robustly identified more reliable and replicable signatures. In agreement with experimental findings, our method identified important microbes of five disease datasets (cirrhosis, obesity, type 2 diabetes, inflammatory bowel disease, and colorectal cancer (CRC). Employing MEGMA and ConvNet, we also discovered highly consistent sets of marker microbes in the cross-cohort CRC patients and microbial shifts in different CRC stages.Funding Information: We appreciate the financial supports from the National Key R&D Program of China, Synthetic Biology Research (2019YFA0905900), Shenzhen Municipal Government grants (No.2019156, JCYJ20170413113448742 and NO.201901), Department of Science and Technology of Guangdong Province (No. 2017B030314083), and Singapore Academic Funds R-148-000- 273-114 and NUS Research Scholarships. Declaration of Interests: The authors declare no competing interests.