Michael Feig
Assistant Professor
- M.S. 1994, Technical University of Berlin
- Ph.D 1999, University of Houston
- Postdoctoral Fellow, 1999-2003, The Scripps Research Institute
- Alfred P.Sloan Fellow, 2005
feig@msu.edu
218 Biochemistry Building
Michigan State University
East Lansing, MI 48824-1319
Office: 517-432-7439
Lab: 517-353-6743
Lab Home Page
Sim DB
Publication search:
Michael Feig Research Interests
We are using computational methods in order to study the structure, dynamics, and energetics of biological macromolecules such as proteins or nucleic acids. Our focus is concentrated in two areas: 1) The modeling of large supramolecular assemblies in atomic detail and 2) the accurate prediction of native protein structures.
As
an example of a supramolecular complex we are investigating the interaction
between the E.coli mismatch recognition protein MutS and DNA
with mismatched or missing base pairs. MutS recognizes defective DNA
after replication and initiates a multi-step process that leads to DNA
repair. A detailed understanding of the DNA mismatch repair system is
relevant for some types of cancer where the repair process is compromised.
Starting from crystal structures of the MutS protein-DNA complex we
apply computer simulation techniques to look at energetic and dynamic
aspects of this system.
The prediction of protein structures from sequence has seen great progress
through
recent methodological advances and the availability of an increasing
number of structural templates from experimental protein structures.
It is now often possible to generate approximate predictions that capture
many of the general features of the native fold.
However, it remains challenging to reach levels of accuracy that are comparable to experimental data unless structures from closely related sequences are available. We are working on new methods that allow the refinement of approximate predictions towards more native-like structures by using new enhanced sampling techniques in combination with an accurate energetic description.
In both cases, the use of an implicit solvent representation is very attractive, since such a model can maintain a realistic description of a biomolecule in aqueous solution without the computational expense of explicit solvent molecules. This allows the modeling of larger systems and/or over longer time scales at a similar level of detail. We are interested in continuing the development of new implicit solvent methodologies in order to improve the level of realism in complex environments.