Protein Folding - a Group Effort

In my last column (“The Problem with Proteins”), I described the problem with protein folding. Today I wanted to describe one of the major efforts underway to solve this problem. And believe me, this really is a group effort. This effort uses a process called Distributed Computing.

Labs at Stanford University, the University of Pittsburg, the University of California at San Francisco, Notre Dame, California State University Long Beach, and the Mediterranean Institute for Life Sciences in Croatia use computers around the world to model and predict protein folding. It is a method of computer processing that runs different parts of a program, or different sets of data for the same part of the program, simultaneously on more than one computer. These individual computers, working at the same time on different parts of the same project, then share their results. This results in much greater computing power than any individual computer can provide.

Okay, so these labs use computers around the world to do the work. What work? What are these computers solving?

Computers are often used to model possible ways for a protein to fold. They can run through the innumerable possible folds and configurations for a given stretch of amino acids in a step-by-step process. At each step, it can analyze how likely it is that the protein folds in that way, based on biophysical and biochemical criteria. Step-by-step, the computer can work its way through from the raw amino acids to what is the most physically likely structure for that sequence. This takes a lot of computing power.

One standard modern computer can simulate how a protein will fold in a nanosecond (that’s 1 billionth of a second) in one day. That’s 1 billionth of a second of folding done in 1 day of work. Unfortunately, proteins fold on an average of around 10,000 nanoseconds. That would take 10,000 days for one computer to analyze a protein from start to finish. Alternatively, if you could link 10,000 computers and have them all run different parts of the prediction, you could finish in 1 day! (I know which method I’d prefer.)

Okay, so by coordinating all of these computers around the world, we can dramatically speed up the predictions and modeling of protein folding. Why is this helpful? Scientists hope that this will allow them to discover the "first principles" of protein folding. Knowing these first principles will allow them to (a) predict what final shape any given protein sequence will ultimately form, (b) make artificial proteins that form a certain shape and thus perform a certain task, and (c) determine why some proteins in our bodies dramatically misfold with dire consequences. There are numerous examples of how misfolding proteins cause human disease, including Huntington's diseases, Alzheimer's disease, and the prion disease Creutzfeldt-Jakob. Figuring out why these diseases are associated with dramatically misfolded proteins is one of the major goals of many protein folding scientists in the world.

If you are interested in checking out the software involved in this project, or even adding your computer to their work, you can read more about them at:
http://folding.stanford.edu/