Search

In response to the need for improving prediction accuracy (ACC) in existing models for Antimicrobial Peptides (AMPs), a novel AMP prediction model called DeepGlap was proposed. This model utilized two protein language models for feature extraction from AMP sequences, followed by fusion of feature vectors. These fused vectors were then input into a deep learning network composed of multiple layers of bidirectional long short-term memory networks (mBi-LSTM), one-dimensional convolutional neural networks (1D-CNN), and attention mechanisms. The model underwent performance evaluation and optimization. Results indicated that the model achieved ACC, the Pearson correlation coefficient (MCC), and the area urder the curve (AUC) values of 0.739, 0.489, and 0.81, respectively, demonstrating superior predictive performance compared to existing AMP prediction models. [ABSTRACT FROM AUTHOR]