Open AccessArticle
NeuroSeq: An Integrated Learning Framework for Enhanced Neuropeptide Prediction and Therapeutic Discovery
by
Rui Wang1,2,#, Yifan Luo2,3,#,Lijiang Huang1, Minfang Zhou1,Yingzi Feng1,Yunlong Zhou2,Yang Wang2,*, Mengpei Zhang1,*
1 Department of Gastroenterology, the Affiliated Xiangshan Hospital of Wenzhou Medical University. Ningbo 315700, China;
2 Wenzhou Institute, University of Chinese Academy of Sciences, Wenzhou 325000, China;
3 The Second Clinical School, Wenzhou Medical University, Wenzhou 325035, China;
# Rui Wang and Yifan Luo contributed equally to this work.
* Authors to whom correspondence should be addressed.
Submission received: May 6, 2025 / Revised: May 14, 2025 / Accepted: May 15, 2025 / Published: May 20, 2025
Abstract
Neuropeptides, crucial in neural signaling and present throughout the nervous system, are integral to processes like stress response and neural repair. Traditional prediction methods fall short due to their complexity, necessitating advanced strategies. Our study introduces NeuroSeq, an integrated learning network that improves neuropeptides' prediction accuracy and efficiency. It employs pre-trained convolutional networks, advanced CNNs, and TCNs to learn complex amino acid patterns automatically. Combining machine and deep learning via average probability voting, an ensemble learning method further enhances predictions. We also propose a technique for optimizing neuropeptide sequences and assessing their reliability, generating high-confidence candidates. Experiments show NeuroSeq surpasses existing models, hastening neuropeptide identification and progressing peptide-based neurotherapeutics. This research emphasizes deep learning's potential in neuropeptide studies, advancing biological insights and therapeutic development. NeuroSeq marks a significant advance in computationally unraveling neuropeptide intricacies, benefiting neuroscience and clinical applications.
Keywords:Deep Learning; Neuropeptide; Machine Learning