My research is in the field of protein structure analysis and prediction, with particular emphasis on modeling appropriate scaffolds for therapeutic drug design. This is part of a broader effort by the structural community on the protein folding problem, and is also tied to attempts at structurally classifying novel sequences pouring out of varied genome projects. The resources of the Computer Graphics Laboratory (CGL) have proved invaluable in rigorously comparing, visualizing and decomposing protein structures in my analysis of the molecular engagements of cytokines and receptors, the structural basis of recognition specificity, and protein structure prediction by packing, topology and symmetry considerations. Structural genomics represents an organized attempt to decode the fold composition of a particular genome, a step beyond the clustering of similar sequences into gene families that may point to related functions. As the universe of sequences map to a smaller number of protein folds, this approach may lead to the broadest evolutionary decomposition of genome information. The next level of analysis is an understanding of how these particular folds interact in three-dimensions, and together form the complex molecular machines that maintain the cell. I am particularly interested in studying signaling pathways that shuttle information from the outside of the cell to the nucleus, and how some of these circuits have been evolutionarily maintained from primitive bacteria to humans. Cytokine-receptor systems have emerged as a central paradigm for molecular recognition due to intense structural efforts -both X-ray and NMR- that have resulted in detailed 3D images of ligands in complex with modular receptors. As cytokines and their receptors fold in a small number of preferred conformations, it has proved useful to focus on the structural invariants of a few paradigmatic interactions and molecules as a means to computationally screen genome databases for further occurrences, and then model them in three dimensions. A simplifying theme in protein structures is their frequent modularization. pathways feature a series of preferred protein-protein interaction domains with conserved structural features that ease their identification from sequence. New modules are often captured by studying internally symmetric sequences that likely represent tandemly repeated modules; catalytic and/or binding sites are then likely located at the interfaces of these modular domains. Fold prediction efforts are eased by the accurate detection of component module structures, and how these domains pack in three dimensions. As an example of this work, an ongoing collaboration with Cynthia Kenyon's group at UCSF aims to structura lly decompose the proteins involved in a primitive insulin signaling pathway in C. elegans.
Showing the most recent 10 out of 508 publications
