Skip to Content

Computational Drug Discovery on Blue Waters with FEP+ software

Maria Argiriadi, AbbVie

Usage Details

Robert Brunner, Erman Guleryuz, Maria Argiriadi, Eric Breinlinger, Kevin Cusack, Michael Friedman, Michael Hoemann, Thomas Vargo, Raymond Huntley, Sami Osman

Protein-ligand binding is an important parameter in 3D computational drug design. There are computationally intense methods called free-energy perturbations (FEP) which calculate the differences in free energies between 2 chemically related molecules in their protein bound and unbound forms. By calculating these energetic differences, binding affinities of these ligands can be predicted/prioritized in a class of compounds against any given drug target. Specifically, Schrodinger FEP+ software (part of the Schrodinger Advanced Small Molecule Drug Discovery Suite) provides a platform to calculate binding free energies robustly and efficiently with the use of parallel GPU computing power. One standard FEP calculation (one pair of ligands, 2 perturbations) takes typically 200 CPU-days, GPU access can make this calculation more tractable (2 perturbations/day on a 4-GPU machine). We currently don’t have access to any high-performance computing resource with GPUs. Access to Blue Waters supercomputer will significantly expedite our FEP workflow and allow us to perform computational drug design at scale. This can potentially transform our drug design approach.

Schrodinger Advanced Small Molecule Drug Discovery Suite is precompiled to run on most commercially supported versions of Linux, including RHEL, CentOS, Ubuntu, and SLES. MPI libraries are provided along with the software, although the use of MPI is optional, and is handled by job control within the software itself. Batch submissions to LSF, SGE, SLURM, and PBS/Torque are supported and a batch queueing system is essential for successful orchestration of FEP jobs. Moab using msub is not supported at the moment, but if jobs are submitted using qsub, integration with Blue Waters is expected to be relatively straightforward.