Folding@home

Protein Folding and Molecular Recognition

Here's a guest post from Prof. Xuhui Huang's lab at Hong Kong University of Science and Technology, another collaborating labortory in the Folding@home consortium.  Prof. Huang and his lab have made several important methodological applications to FAH (for more details, see this review article) as well as important research into the molecular nature of Huntington's Disease.  Here's an update from Prof. Huang:

In the past a couple of years, the FAH has greatly helped us on our research on understanding the mechanisms of the molecular recognition processes. Molecular recognition, such as enzymes need to recognize their substrates and drugs have to be designed to specifically bind to certain receptors, is crucial to biology and medicine.  Experimentally probing the chemical details of molecular recognition events is challenging, while computer simulations have the potential to provide a detailed picture of such events.  With the help of the FAH donors, we are performing large-scale simulations on a group of Periplasmic Binding Proteins aiming to reveal the general relationships between protein structures, its intrinsic dynamics, and mechanism of recognition process. 

The FAH projects related to the above research are between 7700 and 7712.  We greatly appreciate the help from all the FAH donors, beta testers, and the rest of the FAH team to make our research on molecular recognition possible.

 

LTMD: Key new technology for accelerating folding and misfolding simulations in FAH

Here's an update one one of our key projects looking into protein folding, performed in collaboration with Prof. Jesus Izaguirre's lab at Notre Dame.  Below is an update from Prof. Izaguirre on the progress of this project.

The Izaguirre Lab at the University of Notre Dame (http://www.nd.edu/~lcls) has been collaborating with the Pande Lab at Notre Dame to produce a new GPU core that leverages the amazing speed of OpenMM implicit-solvent force calculations (the heart of the GPU core in Folding@home) with new Long Timestep Molecular Dynamics (LTMD). This combination currently allows nearly a 10-fold speedup over OpenMM for systems as small as the WW domain (35 residues, 544 atoms) up to the Lambda repressor (80 residues, 2000 atoms). This translates into about 10 microseconds per day of simulation, which brings single trajectory millisecond simulations closer to FAH. 

In collaboration with Cauldron Development (lead by Joseph Coffland, primary developer of the Folding@home client and also some cores), we hope to produce a GPU core that might be the first hybrid CPU-GPU core. There are technical questions on how to best do this, and we will engage our enthusiastic beta-tester GPU donors to discuss how to best approach this core when we are closer to production mode. 

Going forward, we will continue to improve the LTMD GPU technology to obtain larger speedups for ever larger and biomedically relevant systems. A particularly excitement development will be the extension of LTMD GPU technology to explicit solvent simulations. 

As far as scientific simulations, we are simulating the folding of about 80 mutants of the Pin1 WW domain, a protein implicated in some cancers and Alzheimer's disease. Understanding the role of mutations on misfolding can have important biomedical consequences, since many diseases have at least some component of misfolding of proteins. Another exciting project we are about to start is to simulate the dimerization during folding of proinsulin and proinsulin mutants, which results in some types of Type IA young and adult onset diabetes. 

Thanks to the FAH donors, testers, and to the Pande Lab for their generosity and leadership which has allowed our technological developments and simulations to come this far. 

 

An image of the Pin WW domain.

 

Recent video about Folding@home from the Stanford News Service

The Stanford News Service recently came by to do a video for Folding@home.  Here it is:

2012 Michael and Kate Bárány Award

I have some good news.  Due in large part to the work we've done with Folding@home, I've been named the recipient of the 2012 Michael and Kate Bárány Award for Young Investigators from the Biophysical Society for "developing field-defining and field-changing computational methods to produce leading theoretical models for protein and RNA folding."  This was just annouced in the Biophysics Society newsletter and their web site hasn't been updated just yet for the 2012 awards.

As with all of the accolades coming to our work, this is very much a team effort, and I'm excited that our collective work is being well-recognized.

Results page updated – new key result published in our work in Alzheimer's Disease

We have made a major update to the Results page on the Folding@home web site.  We now list 95 papers that have directly resulted from Folding@home, although we note that there are 193 papers from the Pande group in general.  There are many new results to talk about, but I will just highlight a few below.

One major result is in the area of Alzheimer's Disease (paper #95).  It is believed that Alzheimer's Disease results from the misfolding of the Abeta peptide. Understanding how Abeta misfolds could give us some key insights into how to cure Alzheimer's Disease. This paper experimentally tests a key prediction made in an earlier paper (paper #58: "Simulating oligomerization at experimental concentrations and long timescales: A Markov state model approach" by Nicholas W. Kelley, V. Vishal, Grant A. Krafft, and Vijay S. Pande. J. Chem. Phys. 129, 214707 (2008); DOI:10.1063/1.3010881). In this paper, we show experimentally that there appears to be a beta turn in the Abeta as predicted. This leads to a very stable form of misfolded Abeta which could be used as a starting point for a new Alzheimer's therapy. We are heavily pursuing this research direction at the moment.

Another key result is in the area of methods (paper #93).  We are constantly honing our methods to improve Folding@home's ability to predict the behavior of proteins.  This paper demonstrates the current state of the art in terms of both sampling and analysis.  When compared to detailed TTET experiments, we show that our methods can piece out even fairly detailed aspects of folding.  However, we also see the ways in which our models are not perfect, suggesting how we can improve our methods even further.

We'll highlight other papers as time goes on.  I'm particuarly excited about these results.  In particular, the results from paper #95 were first discovered several years ago, but we carried out several followup studies to verify them, etc.  As always, I'm most excited about what we're doing now and hope to get some of our current key results in theses areas out from peer review soon.

Core 16 for ATI released; also note on NVIDIA GPU support for older boards

In a previous post, I mentioned our plans for supporting ATI GPUs.  I'm happy to announce that with the release of v7 into open beta, we have now released ATI core 16 GPU3 WUs into advanced methods for Folding@home.  Please see Dr. Lin's forum post for details.

We are doing a gradual rollout of these WUs, for a couple of reasons.  First, while this has gone through beta testing, it is still pretty new, so we want to be cautious with its rollout.  Second, I would like to add some more servers to help the load.  These servers are being delivered today and will likley be on line in a week or two.  Depending on the load on the ATI core 16 Work server, we may wait for a full release until the new hardware is on line.

Just a reminder from the previous post that there are some limitations in our ability to support ATI boards.  In particular, we're limited to those which support OpenCL 1.1 in order to get any sort of reasonable performance out of the hardware.  This means that only 5xxx ATI GPU boards or later will work with the new core 16 (the series 3xxx are not supported and the series 4xxx does not have sufficient OpenCL support).  However, we will keep core 11 support going hopefully at least until Sept 1, 2011 in order to support the older hardware.

Also, we are looking at support of older NVIDIA GPUs as well.  In particular, there have been reported issues with some 8xxx series boards.  We may have to limit their support as well due to analogous hardware compatibility issues.

We're excited to be able to release OpenCL support for ATI GPUs.  However, I want to stress that it is still pretty early days for this code, but we're excited about where this is going.

New GPU3 (version 6.41) client released

The latest GPU3 client is now on our high performance client download page.  This new client improves GPU identification and has several other minor improvements.

For those who are unfamiliar with the other FAH clients: there are several types of high performance clients. SMP refers to multi-core CPUs. GPU3 refers to the latest generation Folding@home GPU client. Currently, GPU3 is only available for NVIDIA GPUs, but we are working on a GPU3 release for ATI. For NVIDIA GPUs, we recommend GPU3. For ATI, we suggest GPU2 until GPU3 has passed through its QA testing (going on now). Finally, the v7 Folding@home client (currently in alpha testing) has these functions (SMP and GPU) built in and will not require a special client.

The changes to this version of the client include the addition of a -forcegpu flag ='nvidia_g80_1.0' ; this flag will signal a compute capability of 1.0 (GPU species=10); if the flag ''nvidia_g80' is used the compute capability will be reported as 1.1 (GPU species=11). Also a pop-up dialog box now appears reporting an error, if the -forcegpu flag is unrecognized; the client then exits.

2 new results from FAH

Two new papers from FAH have just come out, both from the lab of FAH affiliate Eric Sorin.  For more information, check out papers #76 and #77 on our results page.

Update in FAH stats by Operating System

We have been tracking down a bug in our stats by operating system page and it looks like we have now found it.  Basically, we were counting SMP clients as giving 1 CPU.  This grossly undercounted the total number of CPUs, especially for Linux and OSX.  Please note that we have only updated this particular page (stats by operating system) and are looking into updating other ones with the additional information of the number of CPUs per client.

There is a remaining known issue with under-reporting the number of  NVIDIA clients.  We are working on this.

For those who are curious, here's the latest as of 5 min ago:

 

Client statistics by OS

OS Type	Native TFLOPS*	x86 TFLOPS*	Active CPUs	Total CPUs
Windows	303	303	291382	3508526
Mac OS X/PowerPC	4	4	4505	141460
Mac OS X/Intel	101	101	24526	134420
Linux	295	295	109289	539508
ATI GPU	896	945	6307	139443
NVIDIA GPU	329	694	2068	215909
PLAYSTATION®3	800	1688	28360	1034788
Total	2728	4030	449343	5714054

Total number of non-Anonymous donators = 1478538

Last updated at Mon, 08 Nov 2010 15:27:38

DB date 2010-11-08 16:38:48

Active CPUS are defined as those which have returned WUs within 50 days. Active GPUs are defined as those which have returned WUs within 10 days (due to the shorter deadlines on GPU WUs). Active PS3's are defined as those which have returned WUs within 15 days.

*TFLOPS is the actual teraflops from the software cores, not the peak values from CPU/GPU/PS3 specs. Please see our main FAQ, FLOPS FAQ, PS3 FAQ,NVIDIA GPU FAQ, or ATI GPU FAQ for more details on specific platforms.

Recent talk of results from Folding@home

Donors are often curious to hear about recent results.  While one can read our papers, listed on our web site, those are fairly technical and intended for a biological or biophysical audience.  The talk I gave at GTC 2010 ("Folding@home:  Petaflops on the cheap today, exaflops soon?") was intended for a computational audience, so it may be more approachable than the papers, especially for those more familiar with the computational side of FAH, rather than the biological.  You can see it here.

I talk a bit about how FAH works and some recent results in protein folding (pushing past the millisecond timescale), protein misfolding disease (Alzheimer's) and viral infection.  If you're curious about how FAH works or what we've done recently, this might be of interest to you.