BiModalAMP_ts

Antimicrobial Peptide Prediction

This webserver deploys the published BiModalAMPts model by Pham et al. for AMP/non-AMP classification. The model was developed on GenPept-Curated-2025, a curated benchmark of 11,000 sequences (1:1 ratio, length 10–200 amino acids).

BiModalAMP_ts integrates:

• Sequence representations learned by a masked Long Short-Term Memory (LSTM)
• 126 quantitative physicochemical descriptors processed by a 1D Convolutional Neural Network (CNN)

Both branches are jointly optimized, and the model is calibrated via Temperature Scaling to support probability-calibrated candidate ranking and efficient CPU inference in resource-constrained screening settings.