SimBioSys Blog

Getting the most out of your computational tools often comes down to proper usage. At SimBioSys our goal has always been to develop tools that are easy to use, with simple interfaces and a lot of automation. However, even though the scripts are simple, as a user you still need to know the best protocol to get the best performance.

Over the past three weeks, we have been working on the SAMPL project (http://sampl.eyesopen.com). SAMPL is a community competition to test protein and ligand modeling. This work has prompted us to formalize the “best practices” protocol for screening with eHiTS and LASSO.

The following graphic illustrates the workflow one should follow when using eHiTS or LASSO for screening databases. The main idea is that we should first try to generate the best model possible, using all available information prior to performing the screen. As with any modeling approach, testing and validating the model is an essential step prior to applying the model in a predictive way. Thus, in this workflow we have a series of “loops” which involve testing and evaluating the model at that step. Before moving to the next step you must be satisfied with the results.

We are looking for comments on this protocol, what do you like, what don’t you agree with, please add your comments. For more information, please visit our eHiTS page.

Over the past few months we have been working on putting the eHiTS docking program on the Cell Broadband Engine (Cell B.E.). Many of you may not have heard of Cell B.E. but you have no doubt heard of Sony’s PS3s. The PS3 uses the Cell B.E. chips to perform very advanced graphical computations mainly for gaming applications. However, the Cell B.E. is one of the most advanced chips on the market and when programmed correctly can offer amazing performance improvements over traditional Intel / AMD processors. SimBioSys is proud that eHiTS is the first major docking program to take advantage of this new advance in computational power, with projected speed-ups of over 40x.
The parallel nature of the Cell B.E. combined with its ability to do vector calculations makes the processor ideal for correctly ported high performance computing applications. It is important to note that getting the full performance out of the the Cell B.E. architecture takes very different programming skills than traditional Intel / AMD processors. SimBioSys has worked closely with IBM and others in the Cell B.E. community to make sure we capitalize on that power. Currently there is a working eHiTS prototype on the Cell B.E. architecture with a planned release in the second quarter of 2008. For more information on Cell B.E. and eHiTS please see the whitepaper found:

http://www.simbiosys.ca/science/white_papers/eHiTS_on_the_Cell.pdf