To address the challenge that traditional single-disease and single-label approaches fail to characterize the complex associations between gut microbiota and multiple diseases, the research team led by Assistant Professor Huang Zhi-An systematically models human gut microbiome-based disease detection as a multilabel classification task for the first time and proposes the GutMLC framework. This method comprehensively leverages the semantic similarity knowledge between diseases and microbes, and introduces matrix factorization and multilabel feature selection, which effectively alleviate the challenges of high dimensionality, sparsity, and heterogeneity in microbial data. Meanwhile, a focal loss function with debiasing weighting is designed, which significantly improves the ability of joint prediction across multiple diseases under extremely imbalanced sample categories. Experimental results demonstrate that GutMLC remarkably enhances the concurrent prediction accuracy of various complex diseases on large-scale cross-project datasets, providing a powerful computational tool for precise diagnosis and treatment based on gut microbiota.

Source：Huang, Z. A., Hu, P., Hu, L., You, Z. H., Chen Tan, K., & Huang, Y. A. (2025). Toward Multilabel Classification for Multiple Disease Prediction Using Gut Microbiota Profiles. IEEE transactions on neural networks and learning systems, 36(7), 12840–12853. https://doi.org/10.1109/TNNLS.2024.3453967