GPU-Accelerated High-Accuracy Molecular Docking Using Guided Differential Evolution
The objective in molecular docking is to determine the best binding mode of two molecules in silico. A common application of molecular docking is in drug discovery where a large number of ligands are docked into a protein to identify potential drug candid
- PDF / 617,140 Bytes
- 19 Pages / 439.36 x 666.15 pts Page_size
- 13 Downloads / 204 Views
Abstract The objective in molecular docking is to determine the best binding mode of two molecules in silico. A common application of molecular docking is in drug discovery where a large number of ligands are docked into a protein to identify potential drug candidates. This is a computationally intensive problem especially if the flexibility of the molecules is taken into account. We show how MolDock, which is a high-accuracy method for flexible molecular docking using a variant of differential evolution, can be parallelised on both CPU and GPU. The methods presented for parallelising the workload result in an average speedup of 3.9 on a four-core CPU and 27.4 on a comparable CUDA-enabled GPU when docking 133 ligands of different sizes. Furthermore, the presented parallelisation schemes are generally applicable and can easily be adapted to other flexible docking methods.
1 Introduction In a modern drug discovery process, in silico identification of ligands (small molecules) which bind to a target protein is often used in the search for novel drug candidates. Such ligands are likely to change the function of the target protein and
This work was done while the author “Martin Simonsen” was affiliated to the Bioinformatics Research Centre (BiRC), Aarhus University, Aarhus, Denmark. M. Simonsen () M.H. Christensen R. Thomsen CLC Bio, Finlandsgade 10–12, Katrinebjerg, DK-8200 Aarhus N, Denmark e-mail: [email protected]; [email protected]; [email protected] C.N.S. Pedersen Bioinformatics Research Centre (BiRC), Aarhus University, C. F. Møllers All´e 8, DK-8000 Aarhus C, Denmark e-mail: [email protected] S. Tsutsui and P. Collet (eds.), Massively Parallel Evolutionary Computation on GPGPUs, Natural Computing Series, DOI 10.1007/978-3-642-37959-8 16, © Springer-Verlag Berlin Heidelberg 2013
349
350
M. Simonsen et al.
may therefore be used to design new pharmaceuticals. From a known 3D structure of a target protein, virtual screening methods can be used to search large ligand databases and create a ranked list of drug candidates. By focusing subsequent in vitro experiments on the top ranked ligands, the cost of experimental testing can be reduced. A common virtual screening approach is to use molecular docking methods to identify binding modes between a ligand and a protein. Molecular docking methods use scoring functions to identify likely binding modes. Rigid docking methods consider only the translation and orientation of the ligand relative to the protein, while the shape of both molecules is fixed during the docking process. In flexible molecular docking the conformation, i.e. the internal structure, of at least one of the molecules is allowed to change. It is common to consider only ligand flexibility while the protein is kept rigid to reduce the number of parameters which have to be optimised. Heuristic methods such as simulated annealing [4, 7], ant colony optimisation [2, 8] and evolutionary algorithms [16, 18] are commonly used to efficiently sample the large search space in flexible docking
Data Loading...
