APIN

APIN identifies antimicrobial peptide (AMP) sequences using an embedding layer and a multi-scale convolutional network (MSCN) combined in a fusion model to improve AMP identification.

Key Features:

• Multi-Scale Convolutional Network (MSCN): Incorporates multiple convolutional layers with varying filter lengths to capture latent features at different scales in peptide sequences.
• Embedding Layer: Transforms raw peptide sequences into dense vector representations that capture essential biological information for downstream learning.
• Fusion Model: Integrates additional biological information by combining multiple sources of data and predictive signals to increase robustness and accuracy.
• Benchmark Performance: Demonstrates superior performance relative to state-of-the-art models on benchmark datasets such as the Antimicrobial Peptide Database (APD)3 and the anti-inflammatory peptides (AIPs) dataset.
• Versatility: Architecture is adaptable to related peptide prediction tasks, including identification of anti-inflammatory peptides (AIPs).

Scientific Applications:

• AMP discovery: Accelerates identification of novel antimicrobial peptides from sequence data.
• Candidate screening for drug development: Provides computational screening of potential AMP candidates for downstream experimental validation.
• Peptide therapeutic research: Facilitates investigation of peptide-based therapeutics, including anti-inflammatory peptide identification.

Methodology:

The method uses an embedding layer to convert peptide sequences into dense vector representations; a multi-scale convolutional network (MSCN) with multiple convolutional layers and varying filter lengths to extract features at different scales; and a fusion model that combines additional biological information and multiple predictive signals.