Global Optimization Challenges in High Resolution Protein Structure Prediction

I will discuss efforts towards transforming structural biology from a largely experimental science to a primarily computational science. Specifically, I will describe

(1) Significant progress in ab initio computation of the structures of small proteins from equence information alone. Using over 100,000 computers distributed around the world, and a physically realistic atomic level forcefield, we can predict structures to less than 2-3Å from the crystal tructure for a number of small proteins (Bradley et al, Science 2005; http://boinc.bakerlab.org/rosetta/rahtoppredictions.php ).

(2) The high accuracy prediction of structures of protein-protein complexes from the structures of the isolated proteins for cases where there are not significant conformational changes, and our progress in docking with backbone flexibility (Furman et al, Science 2005).

(3) The computer based design of novel highly specific endonucleases for targeted genomics applications such as gene therapy (Ashworth et al, Nature in press).

(4) The design of new protein folds and protein interactions, and ongoing efforts to computationally design catalysts for reactions not catalyzed by naturally occurring enzymes and a vaccine for HIV.