(388i) Identifying Ligand Binding Sites and Poses Using Hamiltonian Replica Exchange Molecular Dynamics: Model Systems and a Validation Study Using the Astex Diverse Set
- Conference: AIChE Annual Meeting
- Year: 2013
- Proceeding: 2013 AIChE Annual Meeting
- Group: Engineering Sciences and Fundamentals
- Time: Tuesday, November 5, 2013 - 5:39pm-5:57pm
Fast and accurate identification of protein-ligand binding locations is crucial in drug discovery. Though in theory, MD simulations of such systems will converge on the true bound structure if run long enough, such simulations can get stuck in local minima which leads to insufficient sampling. We report a methodology to accelerate and improve sampling between minima by modifying the Hamiltonian replica exchange molecular dynamics, including using multiple fully coupled and completely uncoupled states as well as Monte Carlo simulation techniques, and using GPU-accelerated code from the OpenMM toolkit.
The methodology was tested on the T4 lysozyme model system, a simple test system with an enclosed, hydrophobic pocket, and validated using 85-system Astex Diverse Set, a well-curated data set with druggable proteins and drug-like ligands, which has been widely used in industrial and academic docking studies. For a 10-ns simulation at each intermediate state for each system, in most cases taking only 1-2 days for each system, all possible binding locations and poses are identified and ranked by statistically analyzing and clustering an ensemble of configurations. Multistate Bennett acceptance ratio (MBAR) is used to calculate the binding free energies of all locations.
In the model system, the true binding site is robustly identified as the top site for three binders, but not a non-binder, demonstrating the power of our methodology to not only identify binding sites, but also differentiate binders from non-binders. Preliminary results for the validation study show that the true binding site is identified as the top three possible sites by our study for about 70% of the cases, with a 38% chance of being the top size. This systematic methodology has potential to substantially help early-stage drug discovery researchers to accurately identify binding sites and poses. The fact that multiple competing binding sites with comparable binding free energies are identified for many cases may provide insights for experimentalists designing drugs for novel targets.