Carlos Simmerling Associate Professor B.A., 1991, University of Illinois at Chicago Ph.D., 1994, University of Illinois at Chicago Postdoctoral Researcher, University of California, San Francisco, 1994-1998. 2000 AMDeC Young Investigator Award. Phone: (631) 632-1336 Email: carlos.simmerling@sunysb.edu Publications Simmerling Group Home Page

Computational Structural Biology

The goals of a computational chemist are to accurately simulate known properties of molecules, assist in the refinement and interpretation of experimental data and predict the results of future experiments. While quantum mechanical methods can be highly accurate, they are limited in that they currently cannot be applied to large systems such as proteins and nucleic acids, and little or no explicit solvent can be included in the calculations. Since the research in my lab involves relatively large biomolecular systems (such as proteins and nucleic acids) where specific interactions with solvent molecules are often important, we use the methods of molecular mechanics. Typical calculations involve molecular dynamics of the molecule of interest along with thousands of explicit solvent molecules, where the behavior of the molecule as a function of time is used to determine kinetic and thermodynamic properties of the system. These simulations can provide an atomic-detail picture of the behavior of a single molecule, rather than the time- and ensemble-averaged views that come from most experiments.

Research Interests

Program Development

One area of current research in my group is the development of new algorithms and programs for accurate and efficient simulation of large biomolecular systems using state-of-the-art computers. I am a member of the development teams for the widely used AMBER and MOIL suites of programs for molecular mechanics calculations. Among the many features of the programs are energy minimization, molecular dynamics, and calculation of free energies. Currently, we are improving the performance of the programs on massively parallel computers, developing efficient genetic algorithms that include solvent effects, evaluating a variety of methods for the inclusion of long-range electrostatic interactions and development of techniques to enhance conformational sampling during simulations of biologically relevant molecules.

Another area of interest in my lab is the development of tools for the visualization and analysis of the large amounts of data that are generated by our calculations. An example of this development is the program MOIL-View for visualization of the structure and dynamics of biomolecules.

Improved Simulation Methodologies: Conformational Sampling

The single largest roadblock to reliable calculations of structures and relative free energies for complex biomolecular systems is the sampling problem. The number of possible conformations for a flexible molecule increases exponentially with the number of rotatable bonds, rapidly exceeding the number which can realistically be evaluated. Overcoming the sampling limitation would have a tremendous impact on our ability to make significant contributions in many areas, such as docking of flexible ligands, refinement of structures with low resolution or incomplete data, quantitative calculation of effects of amino acid mutations on protein stability, assisting in the engineering of modified or new functions for enzymes and catalytic antibodies, and eventually, the "holy grail" of computational structural biology, the prediction of accurate three-dimensional protein structures from only sequence data. The methods that we develop and use must be compatible with the highest quality representations of the system, such as atomic detail, explicit solvation and accurate treatment of the long-range electrostatics that are critical in simulations of highly charged molecules such as DNA and RNA.

Structure Prediction

While the accurate prediction of structures from sequence data alone is a long-term goal, current projects involve the application of new sampling techniques to the study of systems where at least some data is available. Sources of this data include structures of homologous proteins, low-resolution or incomplete experimental data (such as that from X-ray crystallography or NMR spectroscopy), or low-resolution protein structure predictions from methods that forego atomic detail and explicit solvation.

Molecular Recognition

One current application involves prediction of the conformations of antibody hypervariable (antigen binding) loops. The overall structures of different antibodies are conserved despite their ability to recognize and bind diverse antigens, making them the ultimate biological mechanism for molecular recognition. We are developing methods that will predict these structures, including the locations and roles of key water molecules that often mediate antibody-ligand interactions. We also attempt to model and understand the conformational changes (induced fit) that often take place upon antigen binding, and assist in the development and optimization of catalytic antibodies.