BSP-SLIM

BSP-SLIM performs ligand-protein blind docking using low-resolution protein structures to predict ligand binding conformations and prioritize actives for screening.

Key Features:

• Protein structure prediction: Uses I-TASSER to predict protein structures from amino acid sequences for downstream docking.
• Template-driven binding-site identification: Transfers binding-site information from holo-template structures analogous to the I-TASSER predicted models.
• Docking algorithm: Constructs ligand-protein docking conformations by shape and chemical matching into the negative image of identified binding pockets.
• Coarse-grained template approach: Employs a template-based, coarse-grained algorithm tailored to low-resolution or predicted protein structures.
• Benchmark dataset: Evaluated on 71 ligand-protein complexes from the Astex diverse set with an average RMSD of 2.92 Å on binding residues.
• Comparative ligand RMSD: Achieved a median ligand RMSD of 3.99 Å, 5.94 Å lower than AutoDock in the reported benchmark.
• Comparative binding-site accuracy: Reported a median binding-site error of 1.77 Å, outperforming AutoDock by 6.23 Å and LIGSITE(CSC) by 3.43 Å.
• Robustness versus crystal structures: Showed a minimal increase in median ligand RMSD (0.87 Å) and binding-site error (0.69 Å) when using predicted versus crystal protein structures, compared with larger increases reported for AutoDock (ligand RMSD 8.41 Å, binding-site error 7.31 Å) and LIGSITE(CSC) (binding-site error 1.41 Å).
• Virtual screening performance: In virtual screening on six targets, recovered 25% and 50% of actives within the top 9.2% and 17% of the compound library, respectively.

Scientific Applications:

• Blind docking on predicted structures: Predicts ligand binding conformations for proteins with unknown or partially resolved structures using sequence-derived models.
• Virtual screening and prioritization: Ranks and prioritizes active compounds in screening libraries based on predicted docking conformations.
• Structure-based drug discovery with low-resolution models: Enables structure-based docking and screening workflows when only low-resolution or predicted protein structures are available.

Methodology:

Predict protein structures from amino acid sequences using I-TASSER; identify potential ligand binding sites by transferring information from holo-template structures analogous to the predicted models; construct ligand-protein docking conformations by shape and chemical matching into the negative image of the identified binding pockets.