Lambda repressor binding DNA
In the figure, the λ repressor of bacteriophage lambda employs a helix-turn-helix (top; In the figure, the green and yellow ribbons represent the protein's helix-turn-helix domain) to bind DNA (bottom; DNA is highlighted inblue and cyan).
In the figure, the green and yellow ribbons represent the protein's helix-turn-helix domain, while DNA is highlighted in blue/blue-green.
Projects 2106 and 2107
Projects 2106 and 2107 are an extension of projects 2102 and 2103. The previous two were designed to use a flashy new kind of solvent model with power heretofore unknown to mere mortals assigning projects to Folding@home.
Projects 2102 and 2103
Lambda repressor is a simple (5 alpha helices), small (80 amino acids) protein which is often a target for experimental studies of protein folding. Projects 2102 and 2103 are samples of conformations created by holding different combinations of the 5 helices (32 combinations in all) frozen in space while allowing the other helices to move as if they were experiencing 700 K (about 430 degrees C) for 2 ns. [One of the 32 combinations was the whole protein allowed to move at 300 K (30 degrees C)].
More than three hundred conformations were sampled from this library. These conformations will serve a dual purpose: to act as starting points for building a Markov model of protein folding, and to test further the use of Generalized Born implicit solvent models in protein dynamics. It will be especially interesting to see if the use of GB as a solvent model affects the resultant Markov State Model with respect to explicit solvent.
Previous Folding@home work (see this paper) has shown that for solvent viscosities above about 10% that of water, the kinetics scale linearly with viscosity for small proteins: they fold faster at lower viscosities (which isn't surprising) but we can easily correct for the proper viscosity. To test this on a larger protein (lambda repressor) we are simulating at water-like viscosity (project 2103) and at about 10% water viscosity (project 2102).
Projects 2106 and 2107
The current project (2106) is an attempt to discern just how much power is necessary. Both 2106/2107 and 2102/2103 attempt to simulate the presence of water molecules in the trajectory: if we don't have to solve Newton's second law for thousands of water molecules, we can save a lot of time and use that time to solve those equations for the protein atoms: more protein folding gets done. The solvent model used in 2106 is called a "distance-dependent dielectric" which is based on the physical fact that two charges with stuff between them (like water molecules) exert less force on one another than two charges with nothing between (like in vacuum). Thus, the forces between charges (say, between two atoms in the protein) are decreased more with distance than you would predict using the simple form of Coulomb's law found in most introductory physics textbooks.
Note that the time we save by replacing the explicit water molecules with an implicit model will eventually help us to study larger and larger proteins -- big proteins like HIV reverse transcriptase and viruses.
This project, together with 2102/2103, will help us to answer the questions: how much solvent detail do we need, and can we enjoy a simulation speed-up using implicit water while still obtaining physically meaningful results? If the answer to the latter is yes (and surely it is to some degree), then 2106 versus 2102/2103 will help us to answer the question of how much detail we can dispense with: do we need sophisticated implict solvent models, or will simpler ones suffice?Points and deadlines:
p2106 points 202, days 57
p2107 points 404, days 114
|Project Number||Work Unit Name||Number of Atoms||Preferred (days)||Final deadline (days)||Credit||Frames||Code|