A lot of people seem interested in knowing more about the proteins, and kel has done a great job in starting to get people up to date on protein chemistry. I thought I'd throw in a few words to continue the discussion for people who would like a little more background.

As kel mentioned, proteins are amino acid chains - but what are the details? Basically, an amino acid is a chemical moiety, and is the monomer unit of a protein. So, a protein is a polymer of repeating units of amino acids, as illustrated crudely below:

...NH-C(R)-CO=NH-C(R)-CO=[NH-C(R)-CO]=NH-C(R)-CO...

I've enclosed one monomer in [] to clarify. Now the (NH) is the amino part, and the (CO) is the acid part, so that's where the name comes from. The C(R) is my one-dimensional representation of the ALPHA carbon (the C), and the side chain (R) of each amino acid. The R is actually an abbreviation for one of 20 different side chains, but more on that later.

The amino acids are connected in a linear chain as I've represented above, but you'll notice I used 2 different symbols for the connections (also known as covalent bonds): The "-" represents a freely rotatable bond (sp3 for those of you with some chemistry background), whereas the "=" represent a more rigid bond (sp2 type) about which no rotation occurs. So, we know the relative orientation of each CO to the subsequent NH (because there is only one chemically possible orientation and no rotation around the "=" bond). But to fully describe the protein in 3-dimensions, we would also need to know the dihedral angle between each NH-C(R) and the dihedral angle between each C(R)-CO. These are known as the phi and psi angles.

If we know all the phi/psi angles for each amino acid of this protein, then we know its structure (except for the R side-chains, but more on that in another post). Now, naively you might think that these angles could be anywhere from 0-360 degrees, but in reality they are constrained to certain angles, in order to prevent atoms from bumping into each other. You can find out more about these constraints by googling "Ramachandran plot".

So our intial task, for a protein of unknown three-dimensional structure, is to predict the phi and psi angle for every amino acid in the protein. You can see how, as the number of amino acids increases, the problem increases exponentially. This problem was first stated nicely by Levinthal in 1968:

Let's limit each amino acid in a protein to 2 possible conformations (in reality, many more than 2 conformations are possible, as many phi/psi combinations are allowed). For a protein consisting of 100 amino acids, the whole protein chain will have 2exp(100) (ie 2 raised to the power of 100) possible conformations, which is an extremely big number. Only one of these conformations is the right one. Let's imagine the protein can try a different conformation every picosecond. The time required to try all possible conformations would be 10exp(18) seconds (or 10exp(10) years), which is longer than the age of our Universe. Now proteins in our cells clearly find their correct conformation in microseconds to minutes, it doesn't take them billions of years to fold. This is known as the Levithal Paradox.

The analogy to what we are doing, is that even with the help of all your CPUs, we cannot go blindly searching all possible conformations, as this would take too long. Instead, we must be clever with algorithms that only search conformations that are likely to be correct (statistically frequent, chemically reasonable, and many other criteria for "likely to be correct"). But even this subset is large (huge!) which is why your help is so invaluable.

---

Thanks for that explaination! It's going to be tough to beat the people behind this project. You guys are awfully good at explaining all this complex biochemistry. Still, I'm glad I remember a bit from the organic chemistry course I took...let's see when was that? Yikes! Over 30 years ago!

---

-Charlie