NeuroPpred-SVM

NeuroPpred-SVM predicts neuropeptides from sequence data using BERT embeddings combined with sequential features and a support vector machine classifier to distinguish neuropeptide and non-neuropeptide sequences for mechanistic and disease research.

Key Features:

• Embedding-Based Approach: Employs BERT embeddings combined with sequential features to represent complex patterns in neuropeptide sequences.
• Support Vector Machine Classifier: Uses an SVM classifier as a single-model approach that avoids the complexity of multi-layer ensemble models.
• Predictive Performance: Achieves a cross-validation AUROC of 0.969 on training data and an AUROC of 0.966 on independent test sets.
• Comparative Superiority: Outperforms NeuroPIpred, PredNeuroP, NeuroPpred-Fuse, and NeuroPpred-FRL on independent test sets across AUROC, Matthews correlation coefficient, accuracy, and specificity.

Scientific Applications:

• Mechanistic Studies: Accurate identification of neuropeptides supports investigation of their roles in physiological processes.
• Disease Research: Facilitates exploration of neuropeptide associations with diseases and potential therapeutic targets for neurological disorders.

Methodology:

Integrates BERT-based embeddings with additional sequential features to produce enriched sequence representations that are classified by a support vector machine to distinguish neuropeptide and non-neuropeptide sequences.