Stretching proteins and myself into open access (OA)

I'm not sure where to side on the Open Access (OA) publishing business. On the one hand, paying for an article to be published is a regression to the days of page charges albeit without the double-bind that readers are also required to pay. On the other hand, it does flatten access to the article, and often panders to enlightened self-interest by way of increased exposure and citations. Indeed, a strong argument in favor of OA for articles, data and code was just published in the Journal of Chemical Physics by my friend, Dan Gezelter. (Fortunately his Viewpoint is OA and readily available.) Regardless, publishers need to cover their costs, and here lies the challenge to the scientific community. The various agencies supporting science do not appear to be increasing funding to subsidize the fees even while they are making policy decisions to require OA. Libraries love OA because it might potentially lower their skyrocketing journal costs, though no substantial lowering appears to have yet occurred. Long story short, my group is now doing the experiment: We recently submitted and just published our work in PLoS (Public Library of Science.) Props to them for being consistent as they also required us to deposit our data in a public site. I was also impressed by the reviewing process which did not appear to be lowered in any way by the presumed conflict-of-interest that a publisher might have to accept papers (and associated cash) from all submissions. The experiment continues as I'll watch to see how our OA article fares compared to our earlier articles on ASMD in more traditional journals.

Meanwhile, we are excited about the work itself. My students, led by outstanding graduate student, Hailey Bureau, validated our staged approach (called adaptive steered molecular dynamics, ASMD) to characterize the energies for pulling a protein apart. The extra wrinkle lies in the fact that the protein is sitting in a pool of water. That increases the size of the calculation significantly as you have to include the thousands (or more) of extra atoms in the pool. The first piece of good news—that we had also seen earlier—is that ASMD can be run for this system using a reasonable amount of computer time. Even better, we found that we could use a simple (mean-field) model for the water molecules to obtain nearly the same energies and pathways. This was a happy surprise because, for the most part, the atoms (particularly the hydrogens) on the protein appear to orient towards the effective solvent as if the water molecules were actually there.

Fortunately, because of OA, you can easily read the details online. The full reference to the article is: H. R. Bureau, D. Merz Jr., E. Hershkovits, S. Quirk and Rigoberto Hernandez, "Constrained unfolding of a helical peptide: Implicit versus Explicit Solvents," PLoS ONE 10, e0127034 (2015). (doi:10.1371/journal.pone.0127034) I'm also happy to say that It was supported by the National Science Foundation.

Taming the multiplicity of pathways in the restructuring of proteins

The average work to move molecular scale objects closer together or further apart is difficult to predict because the calculation depends on the many other objects in the surroundings. For example, you might find if easy to cross a four-lane street of grid-locked cars (requiring a small amount of work) but nearly impossible to cross a highway with cars speeding by at 65mph (requiring a lot of work.) When the system is a protein whose overall structure is expanded or contracted, there exist a myriad possible configurations which must be included to obtain the average required work. Such a calculation is computationally expensive and likely inefficient. Instead, we have been using a method (steered molecule dynamics) developed by Schulten and coworkers based on Jarzynski's equality. It helps us compute the equilibrium work using (driven) paths far outside of equilibrium. The trouble is that the surroundings get in the way and drive the system along paths that get out of bounds quickly.

In previous work, we tamed these naughty paths by reigning them all in to a tighter region of structures. Unfortunately, this may be too aggressive.  The key is to realize that not all paths are naughty. That is, that there may be more than one region of structures that contribute significantly to the calculation of the work. We found that we could include such not-quite-naughty paths and still maintain the efficiency of our adaptive steered molecular dynamics. I described the early work on ASMD in my recent ACS Webinar.

This research project was truly performed in collaboration. My recent graduate student, Gungor Ozer, did the work while he was a postdoc at Boston University. Tom Keyes hosted him there and gave great insight on the sampling approach. Stephen Quirk, as always, grounded us in the biochemistry.

The title of the article is "Multiple branched adaptive steered molecular dynamics.” The work was funded by the NSF. It was just released at J. Chem. Phys. 141, 064101 (2014). Click on the JCP link to access the article.