In 2013 I switched to protein design. Based on my experience with protein structure prediction I was able to develop a fast, robust and completely computational method for protein design. My work has transformed protein design from a state where <1% designs pass structure prediction (a key check of protein viability) to a state where approximately 50% pass. This improved design strategy has led to a very high success rate with 44 out of 83 proteins ordered having the expected structure. [ ]. This general method will serve as the basis for design of functional de-novo proteins in the near future.
From 2011-2013 I worked on homology modeling. In homology modeling hybridizing homologs outperforms a wider search of conformation space near the homologs. [unpublished work & ]. During these years I implemented the homolog identification and alignment system on http://robetta.bakerlab.org/ and conducted work on how to explore a wide region of conformation space in the vicinity of homologs.
From 2005-2011 I researched de-novo structure prediction and new methods for optimization [ ,, ]. Much of this work remains unpublished due to negligible RMSD improvements(even though energy was significantly improved). As of today only 8% of monomeric proteins < 120 amino acids can be predicted < 3 RMSD [unpublished work].