Folding@home

IMAG meeting today -- The Impact of Modeling on Biomedical Research

I'm at NIH today for the IMAG meeting on the Impact of Modeling on Biomedical Research.  IMAG stands for the Interagency Modeling and Analysis Group and it constitutes all of the science major funding agencies in the US (NIH, NSF, DOE, and more).  It's a great idea to get these agencies together and well-coordinated in terms of their approach to computer simulation and modeling.

In the meeting, they've broken down all of modeling into 5 scales, and I'm chairing the atomic and molecule scale.  Our charge is to talk about the future of modeling and I'll be highlighting some of the key new results from Folding@home.

For more information, please check out:

http://www.imagwiki.org/mediawiki/index.php?title=IFM_Announcement

FAH work highlighted in Biomedical Computing Review

Two key parts of FAH technology -- OpenMM (the software that powers FAH on GPUs) and MSMbuilder (the algorithms that FAH uses to stitch together hundreds of thousands of donor simulations to get coherent results) are highlighted in this months Biomedical Computing Review.  You can download a copy here

http://biomedicalcomputationreview.org/5/4/index.html

Importance of software and data integrity

The scientific code underlying our research has a lot of dials and switches.  It can be very tempting to play with those switches to optimize the calculations.  However, many of the code settings can impact scientific results in non-obvious ways.  It can also hurt scientific reproducibility.  It is therefore extremely important not to modify the scientific cores or data for Folding@Home calculations. Returning any work results with modified cores or data can taint both your results and the results of any WU's calculated on the basis of those.  Thus even well-meaning optimization can potentially hurt the scientific value of the project substantially.  This is why the EULA specifically prohibits such modifications.

When you donate to Folding@Home, we give you a number of "dials" at the client level that you can use to adjust your contributions.  Many donors find that tuning these "dials" and their machine configuration can optimize their contribution.  Please use only the controls provided by the client and the operating system; do not modify the cores or Folding@Home data in any way.  Doing so is the equivalent of providing a tainted donation.  When we detect such tainted donations, we may remove points awarded, decline further donations by blocking work assignment, or take further action.  We greatly appreciate your contributions to Folding@Home; please don't devalue your collective work by undermining the scientific results.

A new type of work unit

We are preparing for the public release of a new work unit category: extra-large advanced methods work units. Some background information is provided below; we'll update this thread with more details as they emerge.

Why a new work unit category?

We have some specific projects where we 1) have large simulation systems and 2) want to get results fast. As multi-core processors get more powerful, we can perform calculations on Folding@Home that

previously required supercomputing clusters.

What's different about these work units from the donor perspective?

These work units are special SMP work units that have larger upload and download sizes, shorter deadlines, and require more memory and CPU resources. That's why we've created a new category.

Is there any points incentive for running these work units?

The base value of these work units corresponds roughly to what an SMP work unit using the A2 core would yield on an equivalent calculation. However, because fast completion is a scientific priority for these work units, we are doing a **trial** of a new bonus scheme where faster WU completion yields a points bonus.

What systems can run these work units?

Right now, only Linux and OS/X systems can run these work units, and they require 8 or more cores. We prefer 8+ *physical* cores, although fast Core i7 machines that are dedicated folders have proven

sufficient during the testing process. The points incentives are designed to match appropriate resources to points value; if your machine is marginal for the extra-large work units, you're probably

better off running standard SMP.

Does this have any relation to the large-points value work units and recent high-scoring users?

Yes. The initial projects are 2681 and 2682, valued at ~25K points base. Although these point values seem high, the work units are correspondingly larger, so the base PPD (points per day) value is

roughly comparable to standard SMP.

A collaborator has donated a large amount of compute time to this project; those clients were initially running under username Anonymous/team 1. To give proper credit for the donation, we have

changed the username to PDC, team 1. During the period of this donation, there are at any time between 100 and 400 8-core clients running under this username (800-3200 cores total).

Please stay tuned for further details regarding the upcoming release.

MD Workshop

Next week, we will be having a Molecular Dynamics (MD) workshop (see http://simbios.stanford.edu/calendar.htm for details) for three main pieces of Folding@home-related software, namely OpenMM (the GPU acceleration code behind the FAH GPU client), MSMbuilder (the software used to parallelize calculations over all of Folding@home, i.e. how to make all the individual donors' calculations work together) and OpenMM/Zephyr (a tool intended to make molecular dynamics simulation easier for non-experts, built from the accelerated OpenMM codebase).

We're very excited about not just making the key software pieces that drive FAH available to other researchers, and moreover, we're working hard to teach others how they work and how to use them efficiently. 

You can learn more about these projects at their simtk.org web pages as well:

OpenMM:  http://simtk.org/home/openmm

MSMBuilder:  https://simtk.org/home/msmbuilder

Zephyr:  https://simtk.org/home/zephyr

Folding@home talk at PARC today, 4pm

Today (January 8, 2008) at 4pm, I'll be giving a talk at PARC, giving an update about Folding@home.  If anyone is local to the Palo Alto area, you can be there in person.  Note that they also do live streaming and should have a video on line (check out http://www.parc.com/forum for details).

UPDATE:  The talk can be found here:

http://www.parc.com/cms/get_article.php?id=799

New paper #61: Computational screen to identify important mutations in influenza

We're happy to announce a new Pande Group paper (paper #61 at http://folding.stanford.edu/English/Papers).  This paper describes a new computational screen to identify important mutations in influenza:

Combining Mutual Information with Structural Analysis to Screen for Functionally Important Residues in Influenza Hemagglutinin. 
Peter M. Kasson and Vijay S. Pande. Pacific Symposium on Biocomputing 14:492-503(2009). 
Download URL: http://psb.stanford.edu/psb-online/proceedings/psb09/kasson.pdf  

The influenza hemagglutinin protein performs several important functions, including attaching the virus to cells it will infect and releasing the viral genome into the interior of the cell. Most protective antibodies against influenza also bind to the hemagglutinin protein. We wish to understand how mutations to hemagglutinin affect viral function, including what keeps avian influenza ("bird flu") from being readily transmissible between humans. In this paper, we have applied a technique from information theory known as mutual information to genetic sequence data to predict important mutation sites on the hemagglutinin protein. In follow-up work, we are combining this technique with other methods to refine these predictions and test some of them using Folding@Home. 

PS For those curious out in more details, check out the paper (see link above) or the technical abstract:

Influenza hemagglutinin mediates both cell-surface binding and cell entry by the virus. Mutations to hemagglutinin are thus critical in determining host species specificity and viral infectivity. Previous approaches have primarily considered point mutations and sequence conservation; here we develop a complementary approach using mutual information to examine concerted mutations. For hemagglutinin, several overlapping selective pressures can cause such concerted mutations, including the host immune response, ligand recognition and host specificity, and functional requirements for pH-induced activation and membrane fusion. Using sequence mutual information as a metric, we extracted clusters of concerted mutation sites and analyzed them in the context of crystallographic data. Comparison of influenza isolates from two subtypes—human H3N2 strains and human and avian H5N1 strains—yielded substantial differences in spatial localization of the clustered residues. We hypothesize that the clusters on the globular head of H3N2 hemagglutinin may relate to antibody recognition (as many protective antibodies are known to bind in that region), while the clusters in common to H3N2 and H5N1 hemagglutinin may indicate shared functional roles. We propose that these shared sites may be particularly fruitful for mutagenesis studies in understanding the infectivity of this common human pathogen. The combination of sequence mutual information and structural analysis thus helps generate novel functional hypotheses that would not be apparent via either method alone.

Visit from Jim Clark and John Hennessy (President of Stanford University)

Jim Clark has been a major donor to Stanford and his great contributions has had a huge impact on my group's work in general and on Folding@home in particular.  Jim Clark was a professor of CS at Stanford, but subsequently was involved in many successful major Silicon Valley companies (see his wikipedia page for all the details).  He also donated over $100M to Stanford to build the Clark Center, a University-wide center for interdisciplinary biology.  

Today, Dr. Clark along with John Hennessy (the President of Stanford University) visited our offices to hear about our recent work.  They both were heavily involved in computer architecture in the past, so they were interested to hear about our work with GPUs and the success we are seeing there (in particular, the significant speed increases).  Also, they are both interested in neuroscience and so I was excited to tell them about our recent Alzheimer's work.

Anyway, I was excited to give them both an update and some idea of where we're going, and it was great fun for me to tell them all about how much we've done.

New paper #60: PS3 paper

We have another paper that came out recently (paper #60 at http://folding.stanford.edu/English/Papers).  This one describes the PS3 client in FAH:

Accelerating Molecular Dynamic Simulation on the Cell processor and PlayStation 3 
by Edgar Luttmann, Daniel L. Ensign, Vishal Vaidyanathan, Mike Houston, Noam Rimon, Jeppe Øland, Guha Jayachandran, Mark Friedrichs, Vijay S. Pande. 

'In this paper, we detail how we were able to get great speed increases for Folding@home (and actually certain molecular dynamics calculations in general) on the PS3. This is our first paper using the PS3, laying out the "how does it work," with a follow up paper in the works describing the results obtained in FAH from PS3 clients. 

It is also worth noting that this paper is a collaboration between FAH team members (Luttmann, Ensign, Vaidyanathan, Houston [now at AMD], Jayachandran, Friedrichs, and Pande) with developers at Sony (Rimon and Øland and their coworkers). 

Here's the more technical abstract as well:

Implementation of molecular dynamics (MD) calculations on novel architectures will vastly increase its power to calculate the physical properties of complex systems. Herein, we detail algorithmic advances developed to accelerate MD simulations on the Cell processor, a commodity processor found in PlayStation 3 (PS3). In particular, we discuss issues regarding memory access versus computation and the types of calculations which are best suited for streaming processors such as the Cell, focusing on implicit solvation models. We conclude with a comparison of improved performance on the PS3's Cell processor over more traditional processors.

Folding@home Alzheimer's Simulation work published

I am very happy to announce that a first key work (paper #58 at http://folding.stanford.edu/English/Papers) from the Folding@home project on Alzheimer's Disease (AD) was just published

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 

url:  http://link.aip.org/link/?JCP/129/214707

Abeta misfolding and aggregation is believed to be the cause of Alzheimer's Disease. Simulations, like Folding@home, are a natural way to understand this process. However, there are several key challenges for simulating the key step -- oligomerization. This work represents a new way to simulate Abeta oligomerization, with a key advance of being able to simulate experimentally relevant timescales and concentrations, using a novel method. We use this new method and the power provided by Folding@home donors to simulate oligomerization in all-atom detail. This has lead to specific predictions about the process, which we are now testing experimentally. 

In many ways, this paper is the "tip of the iceberg" for the Folding@home activities in AD, with a lot more interesting results to come, especially in terms of experimental tests of our predictions and interesting new possibilities for new drugs and AD therapeutics.  So, while we're excited that this result is now past peer review, we're even more excited for what's coming down the pipeline, waiting peer review.  We'll keep you posted as more results become public, hopefully with some even bigger announcements in 2009.

UPDATE:

It was asked which clients participated.  This work started several years ago and took some time to analyze and then publish.  So, it ran exclusively on classic clients.  For the follow up simulations, we are using a mixture of GPU, SMP, and classic clients.  Due to the large number of classic clients, they allow us to calculations not possible on the other platforms.  However, the raw speed (but smaller number) of the GPU and SMP clients allow us to get a good rough idea quickly, refining later with classic clients.