I did not reach my conclusion about [email protected] in a vacuum. I was curious about it, so I asked two bio-chemists about it (without giving them any input from me) and they were both skeptical, and didn't know anyone in bio-tech research who used the data that comes from [email protected] These were very well established bio-chemists who work for major bio-tech firms and have fairly impressive CV's.

Introduction

The [email protected] project (FAH) is dedicated to understanding protein folding, the diseases that result from protein misfolding and aggregation, and novel computational ways to develop new drugs in general. Here, we briefly describe our goals, what we are doing, and some highlights so far.

We feel strongly that a Distributed Computing project must not just run calculations on millions of PC's, but DC projects must produce results, especially in the form of peer reviewed publications, public lectures, and other ways to disseminate the results from FAH to the greater scientific community. Below, we also detail our progress in these areas as well.

Most updates are announced in the main [email protected] blog, but we will periodically update this page. For the latest news, please see the blog.

What is protein folding and how is it related to disease?

Proteins are necklaces of amino acids, long chain molecules.

Proteins are the basis of how biology gets things done. As enzymes, they are the driving force behind all of the biochemical reactions that make biology work. As structural elements, they are the main constituent of our bones, muscles, hair, skin and blood vessels. As antibodies, they recognize invading elements and allow the immune system to get rid of the unwanted invaders. For these reasons, scientists have sequenced the human genome -- the blueprint for all of the proteins in biology -- but how can we understand what these proteins do and how they work?

However, only knowing this sequence tells us little about what the protein does and how it does it. In order to carry out their function (e.g. as enzymes or antibodies), they must take on a particular shape, also known as a "fold." Thus, proteins are truly amazing machines: before they do their work, they assemble themselves! This self-assembly is called "folding."

What happens if proteins don't fold correctly?

Diseases such as Alzheimer's disease, Huntington's disease, cystic fibrosis, BSE (Mad Cow disease), an inherited form of emphysema, and even many cancers are believed to result from protein misfolding. When proteins misfold, they can clump together ("aggregate"). These clumps can often gather in the brain, where they are believed to cause the symptoms of Mad Cow or Alzheimer's disease.

Which diseases or biomedical problems are you currently studying?


Alzheimer's Disease (AD)

AD is caused by the aggregation of relatively small (42 amino acid) proteins, called Abeta peptides. These proteins form aggregates which even in small clumps appear to be toxic to neurons and cause neuronal cell death involved in Alzheimer's Disease and the horrible neurodegenerative consequences.

We have many calculations being performed on AD. Our primary goals are the prediction of AD aggregate structure for rational drug design approaches as well as further insight into how AD aggregates form kinetically (hopefully paving the way for a method to stop the AD aggregate formation).

There have been many projects, including 500 series and 700 series. So far, all of them are either Tinker WUs or normal (not bigWU) Gromacs WUs.

2005

* We are currently in the process of submitting our first paper on FAH results.
* FAH researchers Vishal Vaidyanathan and Nick Kelley present the recent FAH results on AD at BCATS 2005. Their work won the best talk award in 2005.
* Prof. Vijay Pande presented recent FAH work on AD at the National Parkinson's Foundation conference (in the session on AD and its connections to PD).

2006

* Our first paper on AD is ready to submit. We hope to start publicly talking about these results very soon.
* We have submitted our first paper for peer review and we're working on the next 2 paper right now. We're very excited about the results!

2007 We have made some significant progress experimentally testing our computational predictions using NMR.

2008 The first of the papers has come out (see paper #58 on our Results page: "Simulating oligomerization at experimental concentrations and long timescales: A Markov state model approach").

2009 We have had some exciting results regarding new possible drug leads for Alzheimer's. We hope to be submitting these soon for publication.




Huntington's Disease (HD)

HD is caused by the aggregation of a different type of proteins. Some proteins have a repeat of a single amino acid (glutamine, often abbreviated as "Q"). These poly-Q repeats, if long enough, form aggregates which cause HD. We are studying the structure of poly-Q aggregates as well as predicting the pathway by which they form. Similar to AD, these HD studies, if successful, would be useful for rational drug design approaches as well as further insight into how HD aggregates form kinetically (hopefully paving the way for a method to stop the HD aggregate formation).

2006 We are currently in the process of submitting our first paper on FAH results.

2007 Nick has been working on a new collaboration with Judith Frydman's group to computationally test a new hypothesis for HD aggregation found in the Frydman lab.

2008

* Prof. Pande has presented the results on HD at a variety of Stanford internal conferences and meetings. People have been excited and interested in the results.
* We have also started to apply the drug design methods used in Alzheimer's to HD.

2009 New paper #62: The predicted structure of the headpiece of the Huntingtin protein and its implications for Huntington's Disease. It's still early (since this paper was just accepted), but I wanted to give FAH donors a heads up on our work on Huntington's Disease aggregation, which is just about to come out in the Journal of Molecular Biology.



Cancer and P53

Half of all known cancers involve some mutation in p53, the so-called guardian of the cell. P53 is a tumor suppressor which signals for cell death if their DNA gets damaged. If these cells didn't die, their damaged DNA would lead to the strange and unusual growths found in cancer tumors and this growth would continue unchecked, until death. When p53 breaks down and does not fold correctly (or even perhaps if it doesn't fold quickly enough), then DNA damage goes unchecked and one can get cancer. We have been studying specific domains of p53 in order to predict mutations relevant in cancer and to study known cancer related mutants.

2005

* Our first work on cancer has recently been published.
* We are expanding FAH's p53 work to other related p53 systems
* We are getting some interesting results from recent new FAH p53 projects.
* Two new sets of projects have completed and two new papers are being readied for peer-reviewed publication.

2006 FAH researcher Dr. Lillian Chong presented her work on p53 at a lecture at several US Universities.

2007 Plans have started to take a new approach for using FAH to fight cancer: to develop novel chaperonin inhibitors. FAH researcher Del Lucent is taking the lead.

2008 Del has presented his plans to the NIH Nanomedicine center with a very positive response. Planning for the lab side of this work has begun.

2009 Del has been involved in the development of new software methods (Ocker) for the chaperonin inhibitor project.


Osteogensis imperfecta

In collaboration with other groups at Stanford (especially Dr. Teri Klein's group at Stanford University Medical Center), we are looking at Collagen folding and misfolding. Collagen is the most common protein in the body and mutations in collagen leads to a very nasty disease called Osteogenesis Imperfecta (or OI for short). In many cases, OI is lethal and leads to miscarriage. However, 1 in 10,000 people have some sort of mutational in collagen. For many, where the mutation is not very serious, it lies unknown and misdiagnosed and leads to brittle bones and other more subtle problems. In others, however, mutations lead to more serious morphological disorders (as shown on the right).

We are starting to model collagen folding and misfolding in the 1000 series projects. Follow the link for more information.

2005 FAH's first work on collagen has been accepted for publication

2006 FAH researcher Dr. Sangyhun Park presents his work on collagen at a lecture at Duke University

2007 Our paper on collagen folding has been accepted for publication.

2008 Our paper on collagen folding has come out.

For now, our Osteogensis imperfecta stands still as a pilot project, with the bulk of our efforts going into AD and HD.


Parkinson's Disease (PD)

We have also performed preliminary studies on a key protein implicated in Parkinson's disease. Alpha-synuclein is a natively unfolded protein and its folding/misfolding (see figure on the right for misfolded aggregates) appears to be critically linked to PD. We are evaluating the application of various FAH methods to this problem.

2005

* We have only done a pilot study on PD and are looking for funding to continue our work in this area.
* Prof. Vijay Pande presented recent FAH work on AD at the National Parkinson's Foundation conference (in the session on AD and its connections to PD).

For now, PD stands still as a pilot project, with the bulk of our efforts going into AD and HD.


Antibiotics

The Ribosome is an amazing molecular machine and plays a critical role in biology, as it is the machine that synthesizes proteins. Because of this critical role, and some small but fundamental differences in the ribosomes of mammals and bacteria, the ribosome is the target for about half of all known antibiotics. These antibiotics typically work by preventing bacterial ribosomes from making new proteins, thus killing them. We have several projects on going to study the ribosome. Since the ribosome is so huge, these WUs are big WUs and have required us to push the state of the art of FAH calculations. However, with these new bigWUs, FAH is set up to study more and more complex problems, and if successful, with greater and greater biomedical impact.

2005

* We are working on our first paper resulting from FAH's ribosome simulations.
* Prof. Pande presents ribosome results at a protein folding conference at U Penn.
* Prof. Pande presents ribosome results at a lecture at University of California at San Francisco (UCSF) Medical School.
* Prof. Pande presents ribosome results at a lecture at Rice University.

2006

* Prof. Pande presents ribosome result at the NIH Roadmap center on Nanomedicine.
* We are just about to submit our first paper on the ribosome.
* Our first work units for antibiotic drug design calculations are now running on [email protected]

2007 We have received a grant from Stanford University to design and study novel antibiotics. This grant is joint with the labs of Chaitan Khosla at Stanford's Chemistry Department (who does small molecule synthesis, design, and some characterization) and Jody Puglisi at the Stanford Medical School (who studies the ribosome and antibiotics experimentally)

2008 Our first ribosome paper has come out in PNAS. See paper #59. Side-chain recognition and gating in the ribosome exit tunnel.

2009 Our second paper on the ribosome has been submitted for publication.

How are these advances possible?

In order to make breakthroughs using distributed computing, new methods are critical. Distributed computing is an unusual way to perform large-scale calculations. While it gives computer resources much greater than a typical supercomputer (e.g. the almost 200,000 actively processing CPUs in FAH vs. 5,000 in a typical supercomputer), these processors are connected by the Internet, not the high speed, low latency interconnects found in supercomputers. Thus, we must develop new methods to use FAH's unusual computational paradigm and capabilities. Moreover, these methods must be tested.

Much of our work in the first years of FAH has been to develop and test these methods on model systems: small proteins that can be easily studied experimentally. With these experimental comparisons, we can test and validate our methods, as well as find out their limitations (which is critical for improving our methods).

To date, FAH has been very successful, with over 40 published works (as of July 2006) directly stemming from FAH calculations. We will continue to work on all fronts: new scientific cores, new server side algorithms, new models for proteins, and new questions related to testing our methods and applications to disease and other biomedical questions.

Well you certainly were not passing out compliments when you said, "I might have known......"

[Show them this page, and see what they say, please?

Given that Folding is creating untold amounts of green house gases, using up energy supplies that could be put to better uses

m0002a, you keep harping on the associate professor deal, could you offer some more insight into why you think this is such a big thing?

On the impact [email protected] and similar projects has on science, I think they dont really have that much of a direct impact, but when that one lucky computer out there in bum***k Indiana finds the "one" protein that cures/prevents alzheimer, its all gonna be worth it in the end.

[...] Since each gallon of gasoline (just in the gas itself, not counting all the embedded energy to find, extract,transport, refine, etc.) is worth 33.4kWh of electricity; brings the total savings to the equivalent of 21,042kWh -- that is more electricity than our household uses in about 2.5 years!

Next was Wiki where I came upon something known as the Levinthal Paradox, which if I understand it correctly, states that there are 3 to the power of 98 permutations of fold. Given the current 5 petaflops of computation working on folding, I worked out that it would take roughly one thousand million, billion, billion years to go through all the possible permutations.

What kind of science are you participating in, where the trade offs are worthwhile for you?

We took advantage of a generous rebate from our energy company, to greatly reduce the cost of heating our house by insulating it, and after the $2K rebate, our out-of-pocket expense was just $740. We reduced our natural gas bill (during the winter months) from ~$350 by over $200 per month. So, as far as the costs go, we already more than recouped our costs, and the energy company will recoup theirs in a few short years. That is about a 60% energy savings; thank you very much!

Given the current 5 petaflops of computation working on folding, I worked out that it would take roughly one thousand million, billion, billion years to go through all the possible permutations. Given that the age of the universe is only of the order of 14 billion years, I feel that an incredible stroke of luck, to say the least, would be required for [email protected] to give any kind of meaningful return.

Good news - it won't take quite so long. You neglected to consider the rate of change of computational power. If compute power doubles every 24 months [Moore's law], then we should wait for a mere 180 years (90 doubling times) before performing the computation. Our computers will then be 1x10^27 times the current speed, and capable of completing the task in a year.

Actually, it's pretty easy to restrict the conformational space by several orders of magnitudes. One approach is to use either the the amino acid sequence or spectroscopic data to take a pretty good guess at the secondary structure (the internal conformations of the protein) and restrict sampling of the the angles of the backbone to the corresponding allowed angles, called a Ramachandran map. Also you don't need to search the entire conformational space, since you're trying to minimize the energy of your sample structures.
Also, the conformational space of a protein it's a continuous space (it's based on bond distances and bond angles, which can take any decimal number), so in principle there is an infinite number of conformations.

The way it works is that you sample a structure (protein conformation) and you calculate the energy of that structure using an energy function. Usually you do a Monte Carlo simulation and try to minimize the energy of the sample structure, and once you're done minimizing the energy, you should have a nicely folded protein, given that your energy function works properly.

EDIT: Also, I did a protein folding myself last month, which took ~400h on 8 cores, using a spectroscopic consistent energy function (but still very cheap) I've been working on for the past six months. So they age of the universe is, in my opinion, not a completely accurate ballpark. But your point remains valid, it's computationally demanding problem, no question. Hence the need for large CPU resources.

I don't know how complicated this calculation you did was, but from the time frames you describe, I'm sure a complicated calculation makes sense in the context of the [email protected] project !

For someone so concerned about efficient use of energy, he's sure wasting a lot of it in this thread (literally, and figuratively).

There are much bigger fish to fry. I bet at least some of people who stopped folding drive a car for probably thousands of miles a year. If you want to start somewhere, start there.

I think you are missing the point. Very, very few scientists believe that [email protected] is worthwhile, regardless of the energy cost.

Do you have any hard evidence supporting this claim? [email protected] is in the process of getting a paper published that might lead to cures or at least drugs to lesson the effects of Alzheimer's disease. Would this not be beneficial?

Do you have any evidence supporting their claims about [email protected] possibly helping cure Alzheimer's?

It is not my responsibility to prove a negative.

I did ask two respected bio-chemists involved in finding cures for serious diseases, and neither of them saw benefit in [email protected] That doesn't prove anything, but it makes me very suspicious.

If you look at all of the 73 published papers on the [email protected] website, you will find that everyone of them is co-authored by Vijay S. Pande, the person who invented [email protected]

There is very little diversity of scientists involved in using the research being gathered, and the vast majority of them are chemists or computer scientists (not bio-chemists) and mostly associated with the Stanford Chemistry department (not may authors from other universities, and none from bio-tech firms that I could find).

I sure do. Of course, this is all pending peer review.

Very, very few scientists believe that [email protected] is worthwhile, regardless of the energy cost.

If one drives a car, presumably (hopefully) they do it for some reason, such as to get from one place to another.

There is no rational reason to do [email protected] since it is basically just a an erroneous religious belief that it is contributing to curing any diseases.

That does not constitute proof, even in the most remote sense of the word.

Even if by some stretch, a drug was actually developed and went into clinical trials, most clinical trials end without FDA approval.

I never said it did. You simply asked for evidence supporting my claims, not proof. We'll have to wait and see. Clearly the paper hasn't been rejected yet, and with something that has such a high potential impact, I would not be surprised if it spends more than usual time in peer review. Patience is a virtue.