Deep-AFPpred

Deep-AFPpred predicts antifungal peptides (AFPs) within protein sequences using deep learning to accelerate the discovery of novel antifungal candidates.

Key Features:

• Deep learning architecture: A 1D Convolutional Neural Network (1DCNN) combined with Bidirectional Long Short-Term Memory (BiLSTM) is used for sequence classification with transfer learning from pretrained seq2vec embeddings.
• Performance: Reported accuracy is approximately 96% on validation data and 94% on test data, exceeding the performance of other state-of-the-art classifiers.
• In silico screening: Enables computational screening of protein sequences to identify candidate antifungal peptides (AFPs) as an alternative to time-consuming experimental discovery.
• Analytical outputs: Produces predicted AFPs with computed physicochemical properties and motif information for downstream experimental selection.

Scientific Applications:

• Antifungal peptide discovery: Identification of novel AFP candidates for synthesis and experimental testing.
• Therapeutic development: Prioritization of peptide candidates to accelerate development of antifungal therapies.
• Fungal infection research: Support for studies of fungal pathogens, including contexts with increased incidence such as during the COVID-19 pandemic.
• High-throughput screening: Computational triage of large protein datasets to pinpoint potential AFPs.

Methodology:

Deep-AFPpred uses a 1D Convolutional Neural Network combined with Bidirectional Long Short-Term Memory (BiLSTM) and applies transfer learning with pretrained seq2vec embeddings.