Glen Hocky
Undergraduate
University of Chicago, Class of 2009
Major: Chemistry, Math
Research
Charge burial in Biological Systems
I helped prove that the Born model was insufficient for predicting the cost of charge burial in various symmetries relevant to biological problems. This result was published.
Influence of nonlinear electrostatics on transfer energies between liquid phases: Charge burial is far less expensive than Born model. Haipeng Gong, Glen Hocky, and Karl F. Freed. PNAS Volume 105, Number 32. August 2008.
Ab-Initio protein structure prediction and SPEED
Update: This work is featured on the cover of Protein Science, March 2010
Protein structure prediction enhanced with evolutionary diversity: SPEED Joe DeBartolo, Glen Hocky, Mike Wilde, Jinbo Xu, Karl F. Freed and Tobin R. Sosnick. Protein Science, Volume 19, Issue 3, pages 520-534. March 2010. [link].
The Freed and Sosnick groups have developed a highly-sucessful method for Ab-Initio structure predicition. This method involves iteratively running a Monte Carlo Simulated Annealing procedure using a reduced representation where only backbone dihedral angles are sampled. This sampling is done from a fixed torsional library which comes from experimentally derived structures in the PDB.
SPEED -- Structure Prediction Enhanced through Evolutionary Diversity. I have developed a method for using Evolutionary information from a database alignment to increase both the diversity and specificty of the angles that are sampled during simulation. This greatly enhances our Secondary and Tertiary Structure Predictions.
This methodology was used in our predictions for CASP8. This includes our prediction for T0482 (the only true Ab-Initio target) where we achieved the best prediction of 128 groups by RMSD and arguably by GDT (GDT Plot below, full results here).
My work is acknowledged both in our abstract and poster for the CASP8 conference in December 2008. Structure prediction combining the template-based RAPTOR algorithm with the ItFix ab initio method. J. DeBartolo, G. Hocky, F. Zhou, J. Peng, A. Augustyn, A. Adhikar, J. Xu, K. F. Freed, and T. R. Sosnick.
Supercharging OOPS
In collaboration with Mike Wilde (UC Computation Institute, Argonne National Labs), I have been working to run out protein folding code (called OOPS) on a variety of computing resources. Our greatest success has come on Argonne's BGP supercomputer where we have run our code simultaneously on up to 64,000 processors (Incidentally using 4.94 CPU Years in just that one run of ~2 hours).
We wrote up this work, but we still need to submit it somewhere. The report is available from Argonne.
Towards petascale Ab Initio protein folding through parallel scripting. G. Hocky, M. Wilde, J. DeBartolo, M. Hategan, I. Foster, T.R. Sosnick, K.F. Freed
I also conceived and helped write an NSF grant proposal to continue and extend this work. You can get it here.