I-LBR

I-LBR identifies potential ligand-binding residues (LBRs) in proteins using only protein sequence data to support analysis of protein–ligand interactions.

Key Features:

• Query-specific computational methodology: Operates without 3D structural information by constructing models specific to the query sequence.
• Two operational modes (I-LBR^GP and I-LBR^LS): I-LBR^GP provides general-purpose LBR prediction across diverse proteins, while I-LBR^LS performs ligand-specific prediction when ligand type(s) are known.
• Support Vector Machine (SVM) algorithm: Employs a Support Vector Machine trained on a subset derived from the query sequence to predict per-residue LBR probabilities.
• Performance evaluation: Across four testing datasets, I-LBR^LS outperforms I-LBR^GP when ligand-specific information is available, and I-LBR achieves performance that is better or comparable to other state-of-the-art LBR identification methods.

Scientific Applications:

• Protein–ligand interaction analysis: Prediction and analysis of ligand-binding residues from sequence data to support functional annotation of proteins.
• Drug discovery and binding-site characterization: Identification of putative binding sites to inform therapeutic design and target prioritization.

Methodology:

Constructs a training subset from the query sequence, trains a Support Vector Machine (SVM) model on that subset, and predicts the probability of each residue belonging to the LBR class.