Professor
Director, Baker Center
for Bioinformatics and Biological Statistics
Theoretical studies on the structures of proteins, nucleic acids, and
small molecules
Applications of theoretical studies to develop molecular models and
to select new drugs
Computational studies on the structures of proteins, nucleic
acids, and small molecules, and their interactions. Overall the direction
of his research has been to push toward the comprehension of the functions
of large structures. Applications are sometimes made to develop molecular
models and to select new drugs.
Protein Datamining is used to assess protein structures
and their folding patterns. We have evaluated interactions from available
structures and other experimental data. We developed a standard way
to view interaction energies between residues, based on sets of protein
structures. This approach led to useful ways to incorporate structural
and hydrophobicity information into simulations. We also showed that
polar interactions are important when residues come close together;
whereas hydrophobic interactions are effective at longer distances.
New approaches are being developed to associate sequences and structures.
Protein Threading. In application of interaction potentials,
we demonstrated that they are directly useful for selecting the native
forms from among various protein folds.
Conformation Generation. We developed a new approach
to enumerate protein lattice conformations with full efficiency. This
is particularly important for determining native protein conformations,
where the problem is akin to searching for a needle in a haystack, and
random searches are largely ineffective. This new approach opens the
way for the computer generation of much larger numbers of protein conformations.
Libraries of protein-like conformations are being accumulated in which
conformations with secondary structure biases are generated within compact
spaces
Maturation of Hk97 Viral Capsid, passing from immature,
smaller more spherical form to the mature icosahedral expanded form.
Elastic models of Proteins. Large-domain motions of
proteins are computed with simple inter-connected elastic models. These
highly cohesive, cooperative models are most appropriate for considering
the largest functional motions of proteins, which are necessarily independent
of the structural details. Functional mechanisms for processing proteins
or for protein machines can be developed. The methods lend themselves
in straightforward ways to the investigation of the motions of extremely
large biomolecular assemblages of more than 100,000 residues. The results
suggest that high resolution structures are not required in order to
understand the functional motions of proteins. Also, these elastic models
are appropriate for developing pathways for transitions between distinctive
known forms of the same protein. Pathways developed in this way are
more realistic than ones obtained simply by coordinate interpolations,
and can aid in directing atomic simulations along realistic pathways.
Most recent applications have been to the complete ribosome structure;
correlated motions are observed by which the tRNA’s and mRNA pass through
the ribosome structure.
Movie of the ribosome in its slowest motion, showing
a ratchet-like motions where the upper 30S subunit rotates in the opposite
direction to the lower 50S subunit. The bottom view shows the motions
at the interface where the tRNA’s and mRNA are bound.
Systems Biology. We have begun several projects to
integrate the large volumes of data accumulating from genomes, gene
expression, proteomics, and metabolomics by considering various types
of network models.
Machine learning methods are being used to determine
the factors and their relative importance for a given behavior; the
advantage of this approach is that many combinations of factors can
be evaluated rapidly and automatically.
