-- AndrewLeaverFay - 11 Jul 2005

Rosetta adds extra rotamers with explicitly attached water molecules. These rotamers duplicate side chain dihedrals / atom coordinates of some rotamer already present in the rotamer library that has no attached water molecules. These un-solvated rotamers are called ``base rotamers." The explicitly solvated ``rotamers" (solvamers?) are differentiated from the base rotamers by their "amino acid variant" number. The amino-acid-variant number for a base rotamer is either 1 or 2. Histidine is the only amino acid where a base-rotamer has an amino-acid-variant number of 2: it's the histidine ``tautomer" with an AAV of 2. Tautomer (I guess) means an alternate protonation state. For histidine, AAV #2 implies ND1 instead of NE2 is protonated.

I'm not entirely sure how explicit waters work. I think it goes something like this: water molecules are attached to hydrogen bonding groups on the backbone and on to side chains as well. There might be 5 water molecules attached to any particular rotamer. I can imagine one water molecule attaching to the carbonyl oxygen, one to the backbone nitrogen, and maybe a few around the end of a hydroxyl group. Each amino acid variant represents one of the exponential number of water-molecule combinations. With 5 water molecules possibly present for a single base rotamer, there would be an additional 31 amino acid variants.

I believe the energies for these amino acid variants represent the sum of the interactions the individual water molecules have the environment. If so, it is wasteful to calculate the energies of each solvamer separately. Moreover, it's wasteful to represent the solvamer energies seperately.

I think I can refactor this section of the code to see an exponential performance boost. It represents such low-hanging fruit that it's too tempting to put off until later. I'd like to see about getting the code improvements published in the upcoming RECOMB 2006 conference (submission deadline, Sept 26)

Jeffrey Gray at Hopkins uses the explicit water code. His student (an undergraduate) Ryan Harrison wrote me an email last week, interested in trying to incorporate explicit waters into the trie structure. Given it's combinatorial nature, explicit water doesn't belong in the trie. I wrote him to this effect last week, but haven't heard back. I suggest we discuss this idea with Jeff while at RosettaCON^?.
to top