Desmond: High-Performance Molecular Dynamics Simulations for Materials Analysis and Optimization

Molecular dynamics (MD) simulations can provide critical insight into the structure and dynamics associated with materials performance for diverse applications enabling advanced technologies. Condensed phase chemical systems can have equilibration and correlation timescales that are computationally demanding to simulate, posing a practical bottleneck for applying MD techniques for materials simulation in academic, industrial, and government labs. The combination of a highperformance MD code, together with continuously advancing computer hardware technologies, can make atomistic simulations for complex chemical systems tractable in modest time scales, expanding the adoption and impact of MD simulations for materials science. Desmond (created by D. E. Shaw Research) provides an unprecedented combination of parallel scalability, simulation throughput, and scientific accuracy.



Learn more: » http://www.schrodinger.com/Desmond/



For non-commercial users: » http://www.deshawresearch.com/resources_desmond.html

DeepChem - Explore Drug Discovery with Deep Learning

DeepChem is a powerful new open source deep learning framework that offers a feature-rich set of functionality for applying deep learning to problems in drug discovery and cheminformatics.



» Click here to learn more about DeepChem

Accelerating Materials MD Simulation with NVIDIA GPUs

GPU computing has been widely used to accelerate computeintensive simulations and push the boundaries of analysis and discovery. The manycore platform of NVIDIA® Tesla® GPUs provides the computational power far outpacing that of the typical CPU. Scientists and engineers can achieve supercomputer scale performance from their own desktop or HPC/data center to run larger systems, many more systems or for longer simulation timeframes. Exxact Corporation has teamed up with Schrödinger to design a series of GPU computing systems that provide optimum price performance for Desmond users in materials science. The exceptional simulation throughput of Desmond MD running on GPU, and the accuracy of thermophysical property prediction (e.g. glass transition temperature, Tg) using Schrödinger forcefields and MD workflows are illustrated in the Figures below (left and right, respectively).

Figure (Left): Desmond MD throughput on GPU for a system of roughly 25K atoms with the lowest performance showing 175 ns/day using a NVIDIA K20 GPU card (NVT); (Right) Comparison of experimental and DesmondGPU MD calculated glass transition temperatures(Tg) for diverse systems computed using the OPLS2005 forcefield, and the Thermophysical Property module in the Schrödinger MS Suite [40 ns equilibration, 20 K temperature intervals, bilinear interpolation]

Schrödinger MS Suite and Desmond MD Compatible System Highlights

• Designed to meet the requirements for Desmond GPU MD Computing and the Schrödinger Materials Science Suite MD workflows (EOS calculations and thermophysical properties, solubility parameters and explicit mixing simulations, elastic constants, and mechanical properties, etc.)
• Optimized to meet or exceed the published performance numbers
• Preinstalled Desmond GPU computing solutions, designed in collaboration with Schrödinger
• Save time from installing, optimizing, and validating
• Fully customizable to meet your budget