FReMI: a middleware to handle molecular docking simulations of fully-flexible receptor models in HPC environments

摘要

Molecular docking simulations of Fully-Flexible Protein Receptor (FFR) models are coming of age. However, they are computer intensive and their sequential execution can became an unfeasible task. This study presents a middleware, called Flexible Receptor Middleware (FReMI), to assist in faster docking simulations of flexible receptors. FReMI handles intensive tasks and data of totally fully-flexible receptor models in virtual screening and, provides the interoperability between a Web Fully-flexible Docking Workflow (W-FReDoW) and two different High Performance Computing (HPC) environments. FReMI uses internet protocols to communicate with W-FReDoW which helps to reduce the FFR model dimension with a data pattern. Also it sends tasks of docking simulations to execute in a HPC of dedicated cluster and; an alternative model of virtual cluster built on Amazon’s Elastic Compute Cloud (EC2). The results are the FReMI conceptual architecture and two sets of experiments from execution of the FReMI. The first set reports the experiments performed with FReMI using a sample of snapshots from a FFR model on both HPC environments. The second one describes the experiments, on the complete data set, performed with FReMI and W-FReDoW shared execution in a MPI cluster environment on Amazon EC2 instances only. The last set of experiments results shows a reduction of the FFR model dimensionality, transforming it into a Reduced Fully-Flexible Receptor (RFFR) model, by discarding the non-promising conformations generated by W-FReDoW. It also reduces the total execution time to between 10-30% of that of FReMI’s only execution, which, in turn, decreased near 94% with respect to the serial execution.