EXPERIMENT 41 DESCRIPTION:

Experiment 41 involves screening the Enamine library of 2.345 million compounds against the two allosteric sites on the surface of HIV protease. This experiment is similar to Exp. 36, but a different, much larger library of compounds is being screened, and some new targets are being included.

The first part of this experiment involves docking compounds against the two allosteric-fragment-bound crystal structures presented in our recent article in Chemical Biology and Drug Design, vol. 75: 257-268 (March 2010). Three other, brand new crystal structures from Dave Stout's lab that involve allosteric fragments bound to these two sites on the surface of HIV protease are also being used as targets.

Two of these new targets are presented in a new research manuscript from the Stout lab that is currently being peer-reviewed. When this paper is accepted, we will describe these targets in more detail and provide a link to this new paper.


This is by far the largest experiment we have submitted to FightAIDS@Home; it involves faah33,529 - faah45,250 .

The results of these calculations have not yet begun to arrive at TSRI, but your computers should start crunching on them soon.

Right now WU from batch 33,889 are being sent out.

Based on the turnaround time time that we have averaged this week, we should be able to get to the last batch (45,250) in about 710 days , which is around May 8, 2014 .

***** 3 W.C.G. issues that will be affecting how quickly this experiment gets done are:

1. Helf Fight Childhood Cancer (HFCC) has about 70 days to go to finish. Once it is over some crunching power should free up to help FAAH.

2. Computing for Clean Water (C4CW) has about 129 days to go to finish. Once it is over some crunching power should free up to help FAAH.

3. Projects Say No to Schistosoma (SN2S) and Computing for Sustainable Water (CFSW) are currently on MEDIUM Feeder Priority . Once that priority goes to Normal, then it will reduce the free cruching popwer that FAAH benefits from.

Right now WU from batch 34,091 are being sent out.

Based on the turnaround time time that we have averaged this week, we should be able to get to the last batch (45,250) in about 773 days , which is around July 23 , 2014 .

Is it possible or practical to update in the application used for FAAH to create more efficiency, like we have seen with other projects?

How often is the docking application updated to take advantage of new instructions on cpu's and such? It would be nice to see this project accelerated and finished sooner.

Thanks in advance for any input, and sorry if this has been discussed before.

[ot] ryan222h: How often is the docking application updated to take advantage of new instructions on cpu's and such?
I've asked before whether any WCG sciences use any of the x86 vector instructions (MMX/SSE2/AVX etc) at all, but have received no reply.
I seem to recall a forum post several years ago that said they did not do so, mainly because some CPU models might give slightly different results, eg due to the way they handle roundoff of the least significant bit. Not important for graphics, but important for WCG when comparing results in projects that send out 2 copies of each WU.
However, during a beta test of HCMD2, it was found that some AMD CPUs gave slightly different results, associated with the use of SSE2 instructions. [Link]

Although the productivity gains for individual WUs by using CPU vector instructions would be smaller than by using GPUs, the overall productivity of WCG would probably be lifted further by the use of CPU vector instructions, because:
a) it might be applicable to more projects, and at more places in the project algorithms,
b) it eliminates the bottleneck of transferring data to and from the graphics memory - see (a)
c) only a minority of WCG devices have suitable GPUs for GPU projects, or would get them if/when WCG has GPU projects.

Can anyone enlighten us as to how many of the WCG projects currently use CPU vector instructions, and if there are any which do so, how much was gained over using only scalar instructions?

[Edit 17 June]: Hmmm ... the silence on this question is deafening!
[/ot]