Computational Identification of Active Neuropeptides Produced by Human Gut Microbiome with Machine Learning Approaches

Abstract

Materials and Methods: A Long Short-Term Memory (LSTM) model was trained on processed datasets of neuropeptides and non-neuropeptides (amino acid sequences 5-100 in length, divided 80/20 into training and testing datasets). After classifying data extracted from the gut microbiome genome with the developed model, approximately 11 million sequences were predicted as potential neuropeptides. To ensure these sequences were expressed in the human gut, a k-mer analysis was performed, and the results were validated against metaproteomic data. Finally, a smaller, filtered dataset of sequences likely to be expressed in the human gut microbiome was identified and introduced. In the final stage, the filtered sequences were analyzed using the NCBI Protein-BLAST tool to evaluate their similarity to known proteins. As a result, some sequences were identified as potential psychobiotics, though they require further validation to confirm their roles as effective psychobiotics and beneficial microbiomes. Results: This study identified 135 potentially active neuropeptides in the gut-brain axis with 90% accuracy, which were validated using metaproteomic data. These sequences were extracted from the human gut microbiome and introduced as potentially active neuropeptides in the gut. Analysis with the NCBI Protein-BLAST tool showed that some of these sequences resembled beneficial proteins, while others were similar to proteins produced by opportunistic microbes (naughty microbes). With further validation, these sequences could have therapeutic potential. Conclusion: This study demonstrated that deep learning models, combined with precise data preprocessing and bioinformatics analyses, can effectively identify potentially active neuropeptides in the gut-brain axis. The results from NCBI Protein-BLAST revealed that some identified sequences resemble beneficial proteins, while others are similar to proteins from opportunistic microbes. These findings not only provide new insights into the roles of these neuropeptides in the gut-brain axis but also highlight the importance of further investigations to determine their functional roles and therapeutic potential. The identified sequences serve as a foundation for future research, particularly for experimental validation and exploration of their therapeutic effects. However, to practically apply these findings, further studies are needed for biological validation and assessment of these neuropeptides' interactions within biological systems. This computational approach offers a more efficient and scalable alternative to traditional methods such as mass spectrometry and provides new tools for investigating active neuropeptides in the gut-brain axis. Future studies should focus on experimental validation of identified sequences, improving predictive models, and examining the biological roles of these neuropeptides to uncover their full potential in treating neurological, psychological, and gastrointestinal disorders.

Keywords: Machine learning, neuropeptide, psychobiotics, gut-brain axis