SSnet

SSnet predicts protein–ligand interactions by integrating protein secondary structure information (curvature and torsion of protein backbones) into a deep neural network to improve binding-site and interaction inference.

Key Features:

• Secondary Structure Integration: Incorporates curvature and torsion of protein backbones as explicit features to represent secondary structural elements for protein–ligand interaction (PLI) prediction.
• Advanced DNN Framework: Uses deep neural network architectures to process physically informed representations of proteins and ligands.
• Comprehensive Feature Analysis: Visualizes intermediate neural network layers to identify latent spaces and protein features that govern ligand binding.
• Versatile Binding Site Prediction: Identifies primary, allosteric, and cryptic binding sites across multiple protein conformations.
• Non-Specific Molecular Interaction Extraction: Extracts fold-derived structural information to support unbiased prediction of molecular interactions.

Scientific Applications:

• Drug Discovery: Predicts potential binding sites and protein–ligand interactions to support identification of therapeutic compounds.
• Protein Research: Provides latent-space and structural feature insights to inform studies of protein structure–function relationships.
• De Novo Drug Design: Supports design of novel ligands by evaluating secondary structure–based interaction propensities between proteins and candidate compounds.

Methodology:

Uses curvature and torsion data from protein backbones and physically informed ligand descriptors as inputs to a deep neural network and visualizes intermediate layers to extract latent protein features.