A fragment-based approach to the SAMPL3 Challenge.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Fragment-based drug design relies on accurate predictions of binding pose and affinity.
• The SAMPL3 Challenge offered a benchmark to assess fragment linking methodologies.

Purpose of the Study:

• To evaluate a fragment-based design methodology using the SAMPL3 Challenge data.
• To investigate the impact of various modeling aspects on prediction accuracy for protein-ligand interactions.

Main Methods:

• Utilized Grand Canonical Monte Carlo (GC/MC) simulations with annealing of chemical potential.
• Explored protein dielectric models, ligand charge states, solvation energies, and conformational stress.
• GC/MC efficiently identified fragment binding sites by accounting for configurational entropy and bound water.

Main Results:

• Modeling neutral ligands yielded better affinity ranking than charged ligands.
• Inclusion of explicit water molecules in GC/MC simulations improved both affinity and pose predictions.
• Applying ΔΔGs solvation correction further enhanced the ranking accuracy for neutral ligands.

Conclusions:

• Neutral ligand modeling, explicit water, and solvation corrections are key for accurate GC/MC-based affinity predictions.
• The GC/MC method with simulated annealing of chemical potential provides valuable insights into binding affinity prediction parameters.
• This work highlights critical factors for enhancing the success of molecular fragment-based design.