FahCore_11 has reached end of life

FahCore_11 WUs have been out of supply for some time now.  We have been working to see if there are scientific problems which would be well suited by this core, but the science has moved on (FahCore_11 is very, very old and does not support key new advances) and, as mentioned earlier, it is time to retire FahCore_11.

It’s worth mentioning that FAH GPU cores have evolved dramatically in the time since FahCore_11 came out, with Core15 being a major workhorse and Core17 (built on OpenMM) being in major release, and with new cores Core18 (with an updated version of OpenMM) and Core19 (with an updated backend infrastructure) in testing. We’re looking for a lot of exciting new results with these new cores.

A discussion of recent FAH work on protein aggregation-related diseases


About protein aggregation-related diseases

For newly synthesized proteins to become functional, they have to fold into a particular three-dimensional structure or conformation first. During the folding process, a protein goes through a sequence of intermediate states to reach the final functional conformation or the “native state.” Unfortunately, protein folding isn’t fail-proof. Sometimes, proteins misfold and become stuck in certain stable intermediate states without further proceeding to fold into their native states. Such misfolded proteins may aggregate and damage surrounding tissues.

Protein misfolding is implicated in a wide variety of diseases, including Alzheimer’s that affects about half of the population over 85 years of age (1), ALS that claimed the life of the legendary baseball player Lou Gehrig just before his 38th birthday (and leads to all of the recent ALS ice bucket challenges), Mad Cow Disease from eating contaminated beef that leads to spongy lesions in human brains. These diseases manifest different signs and symptoms based on varying factors. Such factors can be the type of misfolded protein and the location in organs that protein aggregation occurs. Some such diseases are limited to one specific organ, some spread to multiple organs; some are inherited, some are acquired; some have known causes, some happen without warnings; some mainly affect certain age groups, some span across generations.

However, these diseases share one trait – they’re currently incurable. Due to the widespread nature of protein aggregation-related diseases and generally poor prospect of treatments, the pathways by which proteins aggregate that contribute to these diseases have become intense subjects of study.

Why Folding@home is well suited to studying protein aggregation-related diseases?

Before we make concrete plans to combat the diseases, we need to know what, when and how it went wrong in the first place. Protein folding is a very dynamic and diverse process where a protein can take thousands of different paths with different conformations to reach its active native state from its initial unfolded state. Numerous folding events can also happen simultaneously. In addition, proteins can be extremely sensitive to small changes of their composing atoms. For example, changing 5 to 10 atoms in each copy of a key protein is enough to make the difference between people who develop early onset Alzheimer’s versus people who don’t get Alzheimer’s at all (2). 

As a result, it’s paramount to capture the entire dynamic folding landscape at atomistic level so that we can pin point and scrutinize the misfolding process. To do so requires enormous computing power – which is where Folding@home comes in.

Design of this study

We analyzed 16 model proteins that had been used in a previous study. They vary significantly in size and folding timescales so that our sample can represent a large protein population. Besides Folding@home, we also included data from the ANTON supercomputer. We adopted the MSM(Markov State Model) approach that has been used to characterize dozens of folding processes, as well as a recently applied method called s-ensemble.

For the purpose of our study, the s-ensemble method works effectively for mainly two reasons. Firstly, s-ensemble is used to study a process similar to protein folding glass forming (3). As a liquid is cooled from high temperature, it may form crystal in which the atoms are arranged in orderly repeating patterns, or it may form glass that lacks such order. Whether the liquid forms one versus the other depends on its chemical properties and ambient conditions. When glass forms, the system pauses at certain stable intermediate states, very much like what could happen during protein folding process. Secondly, among various methods used to analyze glassy state, the s-ensemble method is most reliable as it remains effective when alternative means fail (4).

Major findings of this study

We were able to uncover interesting inactive intermediate states and study their properties at atomistic level. Particularly, these inactive intermediate states are slow-forming (take 10-100μs for smaller proteins, many milliseconds for larger proteins) and long-lived (stable over the course of at least 500 μs). Moreover, they likely emerge from uncommon protein folding pathways. Although their existences are rare events, once they form, they don’t tend to fold into other conformations including the native state. Since such properties of these intermediate states resemble those of intermediate states found in glass, they are referred to as “glassy states” of a protein folding landscape.

            In 7 of the 16 proteins we analyzed, their glassy states contain either all β sheet structures or some different β sheet from the native states. β sheet is a localized region of a protein that looks like a twisted and pleated sheet as a result of a specific bonding interaction among the amino acids that make up the protein chain. The similarities between these β-sheet-rich glassy states and the misfolded conformations of proteins that form toxic aggregates make us speculate that it’s possible for the β-sheet-rich glassy states to seed the protein aggregation process. However, there hasn’t been a unified theory on how aggregation starts, due to the sparseness of supporting experimental data (5,6).

What we can do in the future

Since the glassy states are highly stable and persist over a long time, it offers hope for experimental detection in the future.  In particular, this work shows that perhaps the key essence of misfolding – so critically important for understanding protein misfolding diseases  lies even in the nature of how a single protein folds and misfolds.


(1) “Alzheimer’s Disease Frequently Asked Questions.” New York State Department of Health. Jan 2006. Web. 3 Sep 2014. <https://www.health.ny.gov/diseases/conditions/dementia/alzheimer/alzheimer_qaa.htm>

(2) Paparcone, R., Pires, M., Buehler, M. Mutations Alter the Geometry and Mechanical Properties of Alzheimer’s Aβ (1-40) Amyloid Fibrils. Biochemistry. 2010; 49: 8967-8977.

(3) Bryngelson, J.D., and P.G. Wolynes. 1987. Spin Glasses and the Statistical Mechanics of Protein Folding. Proc. Natl. Acad. Sci. USA. 84:7524-7528.

(4) Jack, R.L., L. O. Hedges, …, D. Chandler. 2011. Preparations and Relaxation of Very Stable Glassy States of a Simulated Liquid. Phys. Rev. Lett. 107:275702.

(5) Dobson, C. M. 2004. Principles of Protein Folding, Misfolding and Aggregation. Semin. Cell Dev. Biol. 15:3 –16.

(6)Luhrs, T., C. Ritter, …, R. Riek. 2005. 3D Structure of Alzheimer’s amyloid-β (1-42) fibrils. Proc. Natl. Acad. Sci. USA. 102:17342-17347.



New servers at Washington University

We (the Bowman lab) have completed our move to Washington University in St Louis and have our new servers up and running.  While our primary server won’t be replacing Folding@home anytime soon, with its 12 cores running at 2.1 GHz, its 64 TB of hard drive space will provide plenty of storage for new projects.  Currently, we’re running a number of projects to understand how rhodopsin detects light and transforms this trigger into an electrical signal that we ultimately perceive as an image.

Recent work from Folding@home highlighted in Biophysical Journal

Our recent work on understanding how protein misfolding occurs (http://www.cell.com/biophysj/abstract/S0006-3495(14)00722-X) has shed light on the nature of misfolding and potential subsequent aggregation (relevant for protein misfolding disease), demonstrating that misfolded states are more prevalent than would be expected, especially due to their metastability (once you get into a misfolded state, it’s really hard to get out of it).

The work was also recently highlighted in a separate article in the Journal (http://www.cell.com/biophysj/abstract/S0006-3495(14)00723-1 ).

New results for “Opa proteins”

We’re excited to share some recent results from our lab that combine simulation and experimental structural biology.  This has been a wonderful collaboration with my colleague Linda Columbus, a Chemistry professor at the University of Virginia.  We are interested in how Neisseria bacteria recognize and infect cells.  This is an important problem #1 because Neisseria are becoming increasingly drug-resistance and #2 because these mechanisms can be borrowed for targeted drug delivery.  Neisseria use a set of proteins called “Opa proteins” on their surface to bind to cells and get inside.  The structure of these proteins is very interesting–the part that sits in the membrane is well-structured, but the part that actually performs recognition is very flexible.  When Prof. Columbus started studying these using NMR spectroscopy (a way to determine molecular structure), the data she got on the recognition end of the protein wasn’t enough to uniquely determine the structure.  My lab and hers partnered to perform molecular simulations of Opa proteins–the recognition part of the protein is indeed flexible, but we were able to use molecular simulation and NMR together to define a bit better how the flexibility works and how it might be related to Opa’s function.  Part of why Opa is so flexible is that it must on the one hand bind to cell receptors but on the other vary enough to evade the human immune response.  We have a theory for what the Opa-cell receptor recognition complex might look like, and we are together performing more simulations and experiments to test this.


The work was published this summer in the Journal of the American Chemical Society:  http://pubs.acs.org/doi/abs/10.1021/ja503093y

Stanford’s Chemistry Department Research Highlights

Recently, for the first time, Stanford’s Chemistry Department did a look back at research highlights from the last (2013-2014) academic year done in the Department.  Folding@home is prominently highlighted:


Stats update back on line

Over the weekend, we had an issue with one our key servers that handles the stats update.  The sysadmins have taken care of it and the stats update is now back on line.

Folding at the Chrome Browsers to Reveal the Secrets Behind the Type II Diabetes

In the past couple of years, Xuhui Huang’s group at HKUST
(http://compbio.ust.hk/) has performed large-scale molecular dynamics
simulations at Folding@Home (Project 2974-2975) to investigate the
mis-folding of the hIAPP (human islet amyloid polypeptide, also called

Like other misfolding peptides, hIAPP is generally unstructured in
water solution but adopts an alpha-helix structure when binds to the
cellular membrane. Around 95% of patients with Type II diabetes
exhibit large deposits of misfolded hIAPP (beta-sheet fibrils).  The
aggregation of this peptide is suggested to induce apoptotic
cell-death in insulin-producing β-cells that may further cause the
development of the type II diabetes.  Using Markov state models
constructed from many molecular dynamics simulations, we have
identified the metastable conformational states of the hIAPP monomer
and the dynamics of transitioning between them.  We show that even
though the overall structure of the hIAPP peptide lacks a dominant
folded structure, there exist a large number of reasonably populated
metastable conformational states.  Among them, a few states containing
substantial amounts of β-hairpin secondary structure and extended
hydrophobic surfaces may further induce the nucleation of hIAPP
aggregation and eventually form the fibrils.  These results were
published at Qin, Bowman, and Huang,  J. Am. Chem. Soc., 135 (43),
16092–16101, (2013) (http://pubs.acs.org/doi/full/10.1021/ja403147m).

In 2014, our lab in collaboration with the Pande group at Stanford
University has successfully developed a new Folding@home client that
can run at the Chrome Web Browsers.  This new core is implemented on
Google Chrome’s Native Client (NaCl) platform (details here:
Currently we have set up a NaCl folding server at Hong Kong
(folding5.ust.hk) to continue our study on the aggregation of the
hIAPP peptides.  Up to now, folding5.ust.hk has collected a few TBs
molecular dynamics simulation data of the hIAPP peptides.

We would like to thank all the donors for their generous
contributions!  We also welcome more clients to try out the new NaCl
Folding@home core.  If you are interested in this new core, you can
download it from the Chrome Store


Bowman lab moves to Washington University

I’ve been an independent researcher at UC Berkeley for the past three years and have now accepted an Assistant Professorship at Washington University in St Louis.  I’ll start the process of building a research team and our computing resources in the next few weeks, so I look forward to starting lots of new projects in the coming academic year!

Folding@home Next Steps Webinar Q&A

Last month Professor Pande gave a webinar/Q&A covering Folding@home’s next steps and accomplishments. Click on the link below to listen to and view the presentation-


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