Update to OS Stats

We’ve taken a look at the stats script that generates the OS Stats page and fixed some under reporting issues we’ve been seeing.  The update script is now live.  The under reporting affected CPU stats (for Win, Lin, and OSX) but not GPU stats.

Progress on connecting computation with experiment

Many biologically relevant conformational changes occur on milliseconds and slower timescales.  Furthermore, many experimental techniques are only sensitive to milliseconds and slower timescales.  Therefore, our ability to reliably capture millisecond timescale events through the use of Folding@home  and Markov state models opens up a host of exciting possibilities.

In one recent study, the Lin, Voelz, and Pande labs teamed up with the Tokmakoff group at the University of Chicago to test the predicted folding mechanism of a protein referred to as NTL9 (here).  The Tokmakoff group specializes in using infrared spectroscopy to probe the details of molecular events, which complements the details we can access through simulation beautifully.  Importantly, these groups were able to demonstrate that a Markov state model for NTL9 correctly predicts details of the protein’s folding mechanism.  This is a great triumph for basic science, and also bodes well for the utility of our (Folding@home) results for proteins that are more closely tied to human disease.

Quantifying structural heterogeneity

As part of Folding@home, you know that proteins change their structures (e.g. transitioning from unfolded to folded conformations).  Once a protein is folded, it is very common to think of it as adopting a single structure though.  To be fair, this is largely because scientists are reasonably good at solving a single, representative structure of a protein but its much harder to assess what sort of structural changes it undergoes with the same resolution.

In a recent paper from my lab, we set out to quantify how much structural heterogeneity there is in folded proteins (here).  We examined a set of proteins ranging from small proteins often used to study protein folding to much larger proteins that are more representative of what is typical in the cell.  We found that there is substantial heterogeneity in every case and demonstrated that our results are consistent with existing experimental data.  Our ability to capture this heterogeneity should be a powerful advantage given the myriad of ways it can affect a protein’s function.

A discussion of recent FAH work on cancer: More technical details

Part II: An In-depth Description of The Study For Readers With a Biomedical Background

A guest post by Jingcheng Wu  

More on c-src: what it is and its relations to cancer

C-src is short for c-src tyrosine kinase. Kinase is a type of enzyme that removes a part (phosphate group) of the molecule (ATP) that is required for every energy-expending process in the body, and attaches it to a specific amino acid (tyrosine, threonine or serine) of a protein (substrate). C-src belongs to a family of kinases called the Src tyrosine kinase.

C-src stimulates the pathways that induce cell growth, generate new blood vessels, prevent cell suicide, and give cells ability to migrate1,2 – all necessary to give rise to proliferation of invasive cancer cells. When there is a mutation to the gene that encodes c-src, mutant c-srcs produced could mimic the functions of the normal signal transduction c-srcs.3 When there is over or mis-expression of the said gene, too many normal c-srcs would be produced. In both cases, it is like stepping on a gas pedal of a car. Once the aforementioned abnormality is coupled with the loss of tumor suppressor gene functions,4 it is like additional loss of the brake of a car, and the car takes off and wreaks havoc.


What is the mechanism of action of existing drugs on the market?

C-src holds the ATP and substrate to close proximity at a small, pocket-like region in itself called the active site. To do so, the c-src active site needs certain shape and atomic arrangement to attract or repulse certain atoms of the ATP and substrate, so that the ATP and substrate fit into the active site of c-src.

Existing small molecule inhibitor drugs like Gleevec (for treating a chronic blood cancer, CML) contain substances structurally similar to ATP.5 Such substances fit into the c-src active site, and disrupt its interaction with actual ATPs.5


Why are existing drugs on the market toxic and ineffective?

The strategy mentioned above is quite effective but has the following flaws:

1: C-src is just one member of one family of kinases, and the active sites of kinases have similar structures because they perform similar functions. Moreover, many kinases do not involve in cancer development and just carry out their benign and necessary tasks in the cells. As a result, such drugs not only inhibit c-src in cancer cells, but also the activities of other friendly kinases in normal cells.

2: Frequent mutations take place in c-src, and some of the mutations can vary the structure of the active site.6 When that happens, ATP might still fit, but the structurally resembling drug substance might not fit.

To tackle this low selectivity and drug resistance problem, we looked at finding a novel drug-binding site on c-src.


What novel drug-binding site on c-src did we find and how does it work?

Proteins including c-src are chains of amino acids. The chains fold, twist, vibrate and change shapes (conformations) all the time due to their internal interactions among the amino acids7 and interactions with surrounding environment.

Inactive c-src undergoes a series of conformational changes to open up its active site allowing substrate binding, and subsequently becomes active. This dynamic process pauses at two metastable conformations before reaching the end active state.16 These two metastable conformations are the intermediates I1 and I2.16 We were able to find a drug-binding site to trap c-src in the I2 conformation so it cannot move on to adopt the fully active conformation. Such task can be achieved by using any drug that binds to a particular region of c-src that is not the active site (hence allosteric site).16 This approach has lower toxicity, fewer side effects and longer lasting action compare to existing treatments on the market.

For the drug-bond region to inhibit c-src conformational change from I2 to active state, it has to communicate with the active site. It doe so via interactions within and between those clustered components of c-src (domain) that behave relatively independently of the rest of the protein. Some of the domains can be many amino acids away from another (long range) yet they communicate with each other and act cooperatively.16 The close-neighboring amino acids also communicate with each other extensively, forming a local network to act in concert to unfold during activation.16


What have we done that had not been done before with c-src?

Before, researchers could tell the static structural differences between active and inactive c-srcs (the two end states). They also identified the metastable intermediates I1 and I2. Yet they did not know how long it took and through what sequence of events and conformational changes to get from one end state to the other. On the other hand, we captured the entire dynamic process. We found out for the first time that the time (106 s) it takes to tansit from inactive state to active state (activation) is about five times longer than the time (21 s) it takes for the reverse process (deactivation). We also discovered the thermodynamic and kinetic features of c-src activation for the first time. Moreover, we found out more details about the two intermediates. In addition, this is the first study that used Markov State Model (MSM) on large-scale complex conformational changes of enzymes, whereas previously MSM had been only used on simulating protein-folding re-arrangements.16 Last but not least, we found all the conformations of the “A loop” (a loop structure critical for c-src activation) of the protein chain, whereas other research groups found only one (which is the open conformation).8


What approaches did we use to conduct our research?

We combined MSM based, massively distributed computational method, statistical method and other algorithms and techniques to simulate c-src dynamic conformational changes at the atomic level.16 There are tens of thousands of atoms in the protein itself and in the surrounding water molecules to simulate. Atoms change their energetic, vibrational and kinetic properties within less than a trillionth of a second. The time it takes for c-src to transit from inactive state to active state is around one tenth of a thousandth of a second,16 which is quite a long time on the atomic scale. The conformational transition also could follow numerous different paths to reach the end states.16 To simulate the entire c-src conformational landscape with surrounding water molecules at the atomic scale, while considering all the possible paths it could take over such a large timescale, it requires enormous computing power and vast amount of resources to carry out.

Luckily, we can break down the entirety of the computation into millions of small parts, and have donors from all over the world to each take one part and complete the computation on their personal electronic devices such as laptops, computers and Playstation3s. The previous studies had to employ simplification strategies that omitted key fine details on the kinetics of conformational transitions due to lack of computing power.


How would the methods outlined in this study potentially increase drug selectivity?

Some kinases, especially the ones in the same family with c-src, have amino acid sequences and structures highly resemble c-src. Such similarities pose difficulties of increasing drug selectivity. Fortunately, our detailed conformational landscapes help to distinguish subtle structural differences among proteins. For example, the Hck kinase (another member of the Src tyrosine kinase family), like c-src, also has two metastable intermediates I1 and I2.9 I1s of Hck and c-src are similar, but the “A loop” of Hck I2 is partially unfolded whereas the “A loop” of c-src I2 is fully unfolded.16 For drugs or fluorescent probes to bind to the novel allosteric binding site, the “A loop” has to be fully unfolded.16 Thus, the drugs or probes would only bind to c-src I2, not Hck I2. In other words, the subtle structural difference detected by our method between c-src I2 and Hck I2 allows drugs to selectively only inhibit c-src.


Other findings from the study

Diversity within intermediate states of c-src: When c-src goes through two intermediates (I1&I2) to switch between active and inactive states, the I1 does not always maintain exactly the same structure, and the same applies to I2. C-src I1 has a partially unfolded “A loop,”16 which is like a loose ribbon flapping around in the wind. In addition, there are a few amino acids of the “A loop” that fluctuate and adopt not two, but multiple intermediate conformations along the activation pathway.16 When taking snapshots of the intermediate states at different times, the “A loop” would look different.


Slow rate of autophosphorylation: As mentioned earlier, c-src is a kinase that transfers a phosphate group (phosphorylation) from ATP to its substrate. It turns out that the substrate of c-src can be another c-src.16 The c-src can phosphorylate a member of its own kind, so this is called trans-autophosphorylation. Activating c-src is like turning on an old-fashioned light bulb. When you first turn on the switch, the light bulb flashes quickly alternating between on (active) and off (inactive) states. The way to lock the light bulb at the “on” state so that you get stable light source is for the c-src to either bind to its substrate (another c-src) or get trans-autophosphorylated by another c-src.10,16 It is important to note that for the substrate to be able to receive a phosphate group from ATP, it has to expose the site that the phosphate group attaches to.9 Initially when there are scarce active c-srcs floating around, the chance of encountering one and subsequently getting trans-autophosphorylated is small.16 As time goes on, more and more active c-srcs are available and the chance of encountering one is much greater, which speeds up the process exponentially.16 Our model predicts that the time evolution of active c-src population is sensitive to changes of concentrations of inactive c-src and briefly active c-src (not phosphorylated so not locked in the active state).


“A loop” has to unfold before C helix can change conformation: The different parts of c-src undergo conformational changes in specific orders. Also, certain parts of the protein like the C helix remains folded during activation,16 although this structure as a whole can rotate inward or outward.


Myristate-binding pocket in c-src could serve as another allosteric drug-binding site: A tyrosine kinase called c-ABL, which is a key component of a mutant fusion protein that causes a chronic blood cancer (CML), has similar fold as c-src.11 C-ABL and c-src both have an ATP binding pocket in the active site, a myristate-binding pocket not in the active site (thus allosteric) and an “A loop.” For c-ABL, binding of myristate to the myristate-binding pocket can be “felt” by the ATP binding pocket and the “A loop.” The ATP binding pocket and the “A loop” respond by changing their conformations, which lead to c-ABL activity supression.12 Due to the structural similarity of c-src with c-ABL, binding of a drug to the c-src myristate-binding site could produce a similar effect as observed in c-ABL.16


Future outlooks based on this study

Future studies on c-src can include its two regulatory domains (SH2, SH3) and its locked active state (trans-autophosphorylated state) in the simulations.16

Furthermore, the discovery of the new allosteric inhibitor drug-binding site can be potentially used simultaneously with the ATP binding pocket of the active site13 as binding sites for a group of drugs called “fragment-based inhibitors.”14 Such a drug has two tightly (covalently) linked fragments that bind to two different sites of the same target kinase.14 It is equivalent to a “super drug” that combines the effect of an existing small molecule inhibitor drug with the effect of a drug that traps c-src in I2.

In addition, a recent crystal structure of CDK2 (a serine/threonine kinase critical for G1 to S phase transition in the cell cycle.17 Inhibitors of it arrest cell cycle and prevent cancer development) is found to be similar to c-src I2.15 A fragment-based inhibitor (AT7519) has been designed to inhibit CDK2.14 Perhaps AT7519 can be slightly modified to inhibit c-src as well, since CDK2 and c-src I2 are structural analogues.

The same methodology and techniques used in this study can be applied to the other members of the Src tyrosine kinase family besides Hck and c-src to find out their subtle structural differences.16 Then these differences can be harnessed for future design of selective drugs that target each individual member like what we did for Hck and c-src.


For more technical details, please refer to the original paper.



  1. Blume-Jensen, P. et al. Oncogenic kinase signaling. Nature. 411, 355-365 (2001).
  2. Wheeler, D., Lida, M. & Dunn, E. The role of Src in solid tumors. Oncologist. 14 (7): 667-678 (2009).
  3. Chial, H. Proto-oncogenes to oncogenes to cancer. Nature Education 1(1):33 (2008).
  4. Knudson, A. Mutations and cancer: statistical study of retinoblastoma. Proc Natl Acad Sci USA. 68(4): 821-823 (1971).
  5. Fausel, C. Targeted chronic myeloid leukemia therapy: Seeking a cure. Am J Health Syst Pharm 64, S9-15 (2007).
  6. Daub, H., Specht, K. & Ullrich, A. Strategies to overcome resistance to targeted protein kinase inhibitors. Nat. Rev. Drug Discov. 3, 1001-1010 (2004).
  7. Bruce, A., Johnson, A., Lewis, J., Raff, M., Roberts, K., & Walters, P. The shape and structures of proteins. Molecular biology of the cell 4th edition (2002).
  8. Meng, Y. & Roux, B. Locking the active conformation of c-Src kinase through the phosphorylation of the activation loop. J. Mol. Biol. 426, 423-435 (2014).
  9. Yang, S., Banavali, N. & Roux, B. Mapping the conformational transition in src activation by cumulating the information from multiple molecular dynamics trajectories. Proc. Natl Acad. Sci. USA. 106, 3776-3781 (2009).
  10. Roskoski, R. Src protein – tyrosine kinase structure and regulation. Biochem. Biophysics. Res. Commun. 324, 1155-1164 (2004).
  11. Cowan-Jacob, S. et al. The crystal structure of a c-src complex in an active conformation suggests possible steps in c-src activation. Structure. 13, 861-871 (2005).
  12. Zhang, J., Yang, P. & Gray, N. Targeting cancer with small molecule kinase inhibitors. Nat. Rev. Cancer. 9, 28-39 (2009).
  13. Martin, M. P. et al. A novel approach to the discovery of small-molecule of cdk2. Chem biochem 13, 2128-2136 (2012).
  14. Gill, A. New lead generation strategies for protein kinase inhibitors-fragment based screening approaches. Mini Rev. Med. Chem. 4, 301-311 (2004).
  15. Betzi, S. et al. Discovery of a potential allosteric ligand binding site in cdk2. ACS Chem. Biol. 6, 492 (2011).
  16. Shukla, D., Meng, Y., Roux, B. & Pande, V. Activation pathway of Src kinase reveals intermediate states as targets for drug design. Nat. Commun. 5:3397 doi: 10.1038/ncomms4397 (2014).
  17. Tsai, L., Lees, E., Faha, B., Harlow, E. & Riabowol, K. The cdk2 kinase is required for the G1-to-S transition in mammalian cells. Oncogene. 8(6): 1593-

A discussion of recent FAH work on cancer: A brief overview

Part I: A Brief Overview of The Study for The General Public

Guest post from Jingcheng Wu.
Cancer affects the general population in an extensive and intensive way. It accounts for 1 out of 4 deaths in the US.1 The global annual cancer cases are expected to rise to 22 million within the next two decades.2 Existing drugs used in chemotherapy on the market are not only ineffective 97% of the time,3,4 but also cause severe damage to the body as a whole due to the drugs’ high toxicity. We are all too familiar with the frightening adverse effects of chemotherapy such as hair-loss, holes in intestine, swelling of the body, feeling sick and tired, abnormal bleeding, to name a few. Many cancer patients choose death over going through the agony of chemotherapy by refusing the treatments.

The reason behind the severe toxicity of anti-cancer drugs lies in their low selectivity. Aiming at killing cancer cells, the drugs also massively destroy normal cells and impede the growth of new healthy cells. Thus come the tragic sufferings very often seen in the oncology wards. The current cancer drugs attempt to cure the patients while kill them at the same time. The solution, then, lies in finding a new way of targeting cancer cells with minimal harm to normal cells.

An ideal target for cancer treatments should be commonly found in a wide range of cancer cells and exists in far less amount in normal cells. Researchers found such a target, which is a protein called the “c-src.” Among half of the most common and most lethal human cancers, high c-src activity has been detected,5 and is found to be an essential element that causes cancer. Therefore, inhibiting c-src activity can contribute to a key breakthrough in cancer treatment, and will in turn benefit a potentially large world population that may be devastated by c-src dysfunction.

An ideal drug in this case should only inhibit c-src activity without interfering with normal cellular functions carried out by other proteins, including the ones that highly resemble c-src. As the famous quote from Art of War goes, “if you know both your enemy and yourself, you need not fear the results of a hundred battles. If you only know yourself but not the enemy, for every victory gained you will also suffer a defeat,” it is apparent that having a thorough understanding of c-src is key to designing the ideal drug, so that we can find out its potential “weak spots” and get a grip on them. It is an easy realization in theory but hard implementation in practice due to the immense computing power required to scrutinize the dynamic behavior and structure of c-src. All we had before were a few static snapshots of it. A detailed dynamic picture of c-src had not been available until Folding@Home6 came into the picture.

In order to do so, Folding@Home harness the unused computing power in personal electronic devices from volunteers worldwide. The combined computing power makes Folding@Home computing network the fastest super computer in the world.7 As a result of that collective effort, we simulated how each atom of c-src and surrounding water molecules moves, and we discovered a major “weak spot” of c-src that can be exploited to suppress c-src’s cancer-causing activities.

This weak spot exists exclusively in an intermediate stage in c-src’s transformation between inactive and active states. These details on c-src structural changes in various states are difficult to study; yet they turn out to be very important to developing a drug with the desired characteristics. (Refer to Part II for details.) The newly discovered “weak spot” is a unique structure on c-src that binds to certain drugs. It makes an ideal drug possible by allowing certain drugs to only influence c-src but nothing else, which minimizes damage to normal cells.

This study is a fantastic starting point and template for future studies to build onto. We can add more complex interactions to future simulations. Also, multiple structures on c-src can be used together for a single drug to exert high potency. In addition, the same techniques can be used to study other cancer-related proteins at atomic level, and in turn their subtle structural differences can be used for future drug design with high specificity and selectivity.

The Folding@Home mission is a beautiful example of the world uniting to combat a common enemy of humanity. The collective power of our research group and our donors push the frontiers of biotechnology to new limits and redefine the impossible.



  1. American Cancer Society. 2014 Cancer Facts and Figures.
  2. World Health Organization. Cancer Key Facts.
  3. Cairns, J. The treatment of diseases and the war against cancer. Scientific America. 253(5): 51-59 (1985).
  4. Morgan, G., Wardy, R. & Bartonz, M. The contribution of cytotoxic chemotherapy to 5-year survival in adult malignancies. Clinical Oncology. 16: 549-560 (2004).
  5. Dehm, S. & Bonham, K. SRC gene expression in human cancer: the role of transcriptional activation. Biochem. Cell Biol. 82 (2): 263–74 (2004).
  6. Shirts, M. & Pande, V. Screen savers of the world unite!. Science. 290, 1903-1904 (2000).
  7. PS3 network enters record books. BBC News. 02 Nov 2007. Web. 15 Mar 2014. <http://news.bbc.co.uk/2/hi/7074547.stm>.





New team member: meet Jingcheng Wu

Ms. Jingcheng Wu is a new team member, whose work is to help improve communications between FAH and its donors, especially explaining (in a non-technical way) what FAH has been able to do.   Ms. Wu was born and raised in China. She moved to the US as an exchange student at age 16 and has been living in numerous parts across the US. She is a past medical student with experience in a wide range of fields such as scientific research, healthcare, education, international business, media, journalism, retail, investment banking, and performing arts with proven success. She graduated Summa Cum Laude with a BS in chemistry from the George Washington University. Her multi-cultural background, interest in medicine and passion to make a positive influence drove her to serve as an intern in the Pande group and contribute to their mission. She is currently living on Stanford University campus.


Major Milestone for FAH: breaking the 40-PetaFLOP barrier

We’d like to thank and congratulate all of our donors for helping us break the 40-PetaFLOP barrier.  We’ve been working hard to make it easier to run the client and also making more of a push to get the word out about what FAH has done and what it’s doing.   It’s great to see this response.

We’re also grateful to some large corporate donors.  At the moment, they would like to stay anonymous, but we are expecting to make an announcement about their participation some time in the future.  We’ve also been working on new directions for FAH, with some of them recently released (such as the NaCl client) and others still in the works.

Finally, I’d like to note that we have a strategic plan for the types of new calculations we can do at the 100 PetaFLOP and 1000 PetaFLOP scale.  When/if we reach those levels, we’re excited to roll out those new, even more ambitious projects.

The release of our latest Folding@home desktop client

Here is a guest post by Joseph Coffland, Folding@home Developer:

We are very happy to announce the release of our latest Folding@home desktop client. This version sports a new streamlined Web interface as well as numerous improvements to the user experience and folding performance. Both our internal and beta testing teams have worked hard to ensure that this is one of the best tested and most stable Folding@home clients to date.


The new Web interface was designed to be easier than ever to use. It now displays your user and team points and has a very clear start and stop button and a power slider bar which allows you to quickly control how much of your computing resources are contributed. Power users can continue to use the FAHControl advanced interface which enables monitoring and control of entire folding farms.

Of course Folding@home would not be possible if not for its contributors. The Folding@home network currently consists of about a quarter of a million active computers and is nearing a top speed of 40 PetaFLOPs. That is faster than any scientific computing system in the world. Although these achievements are impressive, we are aiming to push the envelope even further with innovations such as the recently released NaCl folding client which allows you to run Folding@home in the Google Chrome browser without installing any additional software. Upcoming innovations in this area will make it possible to rapidly grow your folding teams using social media.

To achieve both the technological and scientific goals we’ve aimed to make this client easier to use and more reliable than ever so you can feel confident about encouraging others to install Folding@home. If you do run into problems our ears are open. Join the discussion at foldingforum.org where numerous Folding@home experts are ready to help. Happy folding!

Joseph Coffland
Folding@home Developer
Cauldron Development LLC

Working to fix the shortage of GPU WUs

We’ve had a large influx of GPU clients, including a big donation of time from a corporate partner, and we’ve run a bit low on GPUs late last night (pacific time).  The team got new WUs going last night and we expect they should be online shortly.

While this shortage is naturally annoying for donors (hopefully just briefly annoying, with the new WUs coming online soon), there is an interesting upside to this––we’ve been really blown away with all the GPU resources donors have been running and look forward to the exciting research we can accomplish with this great outpouring of support.

Donate Bitcoins to Folding@home

We’re grateful for all of the donations we get. Donors have often asked us for other ways to donate and we wanted to give a heads up on a new way to do so. You can now help support Folding@home by donating Bitcoins. Click on
Donate Bitcoins
in the “Help us reach 1,000,000!” tab. Enter any bitcoin amount (whole or fraction) you want into the window. You can also help support our research by donating using your credit card.

Add your computer's power to over 327,000 others that are helping us find cures to Alzheimer's, Huntington's, Parkinson's and many cancers ...

... in just 5 minutes.

Step 1.

Download protein folding simulation software called



Step 2.

Run the installation. The software will automatically start up and open a web browser with your control panel.

Step 3.

Follow the instructions to Start Folding.

Stanford University

will send your computer a folding problem to solve. When your first job is completed, your computer will swap the results for a new job.

Download the protein folding simulation software that fits your machine.


Installation guide