NeuroPred-FRL: an interpretable prediction model for identifying neuropeptide using feature representation learning.

Area of Science:

• Immunoinformatics
• Computational Biology
• Drug Discovery

Background:

• Neuropeptides (NPs) are key neurotransmitters in immune regulation.
• Accurate identification of NPs is crucial for research and therapeutic development.
• Existing NP prediction tools require performance enhancement.

Purpose of the Study:

• To develop an advanced machine learning-based meta-predictor for neuropeptide identification.
• To improve the accuracy and efficiency of large-scale NP prediction.
• To facilitate immunoinformatics research and accelerate drug development.

Main Methods:

• Developed NeuroPred-FRL using a feature representation learning approach.
• Generated 66 baseline models using diverse encodings and classifiers with feature selection.
• Constructed the meta-model by optimizing probability scores and employing a random forest classifier.

Main Results:

• NeuroPred-FRL demonstrated superior prediction performance compared to state-of-the-art predictors.
• Benchmarking via cross-validation and independent tests confirmed high accuracy.
• Model interpretability was achieved using the SHapley Additive exPlanation algorithm.

Conclusions:

• NeuroPred-FRL is a powerful tool for large-scale neuropeptide identification.
• The predictor facilitates understanding NP functions and their clinical applications.
• This work advances immunoinformatics and supports therapeutic strategies.