FastROCS TK

FastROCS TK

Real-Time Shape Similarity for Virtual Screening, Lead Hopping & Shape Clustering

While ROCS is a fast tool for shape-based scoring, able to process up to 100 molecules per CPU second, there are some important tasks that cannot be solved at this speed including searching virtual libraries of billions of compounds or clustering millions of molecules by shape. Also at this speed shape similarity is not an interactive task. To enable both of these new applications our shape similarity technology we have recently ported ROCS to the GPU; this tool is called FastROCS. The impact of the application of GPU technology to shape searching is immediately apparent from Figure 1; the speed increase on going from CPU to GPU is over 3 orders of magnitude. The capability of commodity GPU units continues to improve, with speed increasing over 400% from 2011 to 2016, and it is anticipated that this rate of improvement will continue for some time, resulting in FastROCS becoming proportionately even faster than the traditional CPU version.

FastROCS now performs shape searching at speeds hitherto only realized by more approximate 2D methods such as fingerprints, allowing much more physically realistic shape similarity calculations to be performed on large databases of millions of molecules in interactive time (10 seconds or less).

Note that FastROCS is available both in toolkit form and as a web-based graphical user interface. For more detailed information on FastROCS TK click the documentation link below (be sure to select your programming language and then choose "FastROCS Toolkit" from the list on the left) or contact us to evaluate.


 Documentation   >   Evaluate
Figure 1: Acceleration of shape searching on the GPU.

Figure 1: Acceleration of shape searching on the GPU.

 illustration of the enabling power of FastROCS

Figure 2: A recent illustration of the enabling power of FastROCS comes from Boehringer-Ingelheim (BI) and their proprietary virtual chemistry database of around 1*10^9 molecules, BICLAIM. BI have used ROCS to search BICLAIM for some time, with significant success.1 However the searches are resource intensive (requiring over 11 CPU years, or around 2 days on a 2000 CPU cluster, to complete ) requiring a substantial investment in hardware to ensure that the searches complete in a reasonable time. Performing the same searches with FastROCS requires only around 5 GPU days (see Figure 2), allowing them to be performed on a single 4 GPU machine in a single day.

Features

  • Processes millions of conformations per second
  • Returns overlays based on the quality of the 3D shape match against the query
  • Overlays are intuitive and visually informative when viewed in standard visualizers (e.g. VIDA)
  • Available as an XML-RPC based web service
  • Jobs can be launched and the subsequent results viewed directly from within VIDA
  • Reports rigorous Tanimoto measure between shapes

Hardware

OpenEye can assist in acquiring hardware pre-installed with FastROCS. We collaborate closely with Exxact to deliver a machine tuned for the highest level of performance. A wide selection of hardware is available here. Please don’t hesitate to contact sales@eyesopen.com for more information.

References

  1. A fast method of molecular shape comparison: A simple application of a Gaussian description of molecular shape Grant, J.A., Gallardo, M.A., Pickup, B., J. Comp. Chem., 1996, 17, 1653.
  2. A shape-based 3-D scaffold hopping method and its application to a bacterial protein-protein interactionRush, T.S., Grant, J.A., Mosyak, L., Nicholls, A., J. Med. Chem., 2005, 48, 1489.
  3. Comparison of Shape-Matching and Docking as Virtual Screening Tools Hawkins, P.C.D., Skillman, A.G., Nicholls, A., J. Med. Chem., 2007, 50, 74.
  4. Assessment of Scaffold Hopping Efficiency by Use of Molecular Interaction Fingerprints Venhorst, J., Nunez, S., Terpstra, J.W., Kruse, C.G., J. Med. Chem., 2008, 51, 3222.
  5. Multiple protein structures and multiple ligands: effects on the apparent goodness of virtual screening resultsSheridan, R.P., McGaughey, G.B., Cornell, W.D., J. Comput. Aided Mol. Des., 2008, 22, 257.
  6. Lessons in molecular recognition. 2. Assessing and improving cross-docking accuracy Sutherland, J.J., Nandigam, R.K., Erickson, J.A., Vieth, M. J. Chem., Inf. Model, 2007, 47, 2293.