Hadoop Spark Based Hydrogen Bond Analysis Tool (H-BAT) for Molecular Dynamics Simulation Trajectory Data

Abstract

Molecular dynamics (MD) is a computational technique that works on the Newton's equations of motion to study the dynamics of various biomolecules and, is commonly used by structural biologists. With the development of advanced simulation techniques and increasing computing power, large amounts of data are being generated from these simulations. Various enhanced sampling techniques are currently being used, that are able to capture rare events and generate simulation data in the form of multiple trajectories. Analyzing the simulation trajectory data and extracting meaningful information using the traditional sequential post-simulation data analysis methods are becoming increasingly untenable. Currently, molecular dynamics simulation algorithms that are scalable on high-performance computing clusters are available which generate a huge amount of MD data in short span of time. The need of the hour lies in developing a advanced and high-performance analytics platform based tool that can analyze this huge simulation data in a faster and more efficient way. The Hadoop Spark framework, provides an excellent platform that meets these requirements of handling large amounts of data parallely and perform analytics with high scalability. In this study, a tool name H-BAT has been developed using the Hadoop Spark platform to calculate hydrogen bonding within all solute-solute, solute-solvent and solvent-solvent molecules in large MD simulation trajectories. Vector geometry has been used for calculation of angle and distance between the atoms which are present in the form of triplets of filtered atoms taking part in hydrogen bond formation. The benchmarking was performed up to a data size of 48 GB which showed linear scalability. Additionally, the tool is capable of handling multiple similar trajectories simultaneously. Future enhancement of the tool would include various other analysis like normal mode analysis, RMSD, 2DRMSD and Water Density Analysis using the Hadoop Spark framework.<br>