||Proteins are chemically simple molecules, being unbranched polymers of uncomplicated
organic compounds. Nonetheless, they fold up into a dazzling variety of complex and
beautiful configurations with a dizzying array of structural, regulatory, and catalytic
functions. Despite great progress, we still have very limited ability to predict the
folded conformation of an amino acid sequence, and limited understanding of its dynamics
and motions. Thus, this work presents a quartet of interrelated studies that address
some aspects of the detailed local conformations and motions of protein backbone.
First, I used a density-dependent smoothing algorithm and a high-quality, B-filtered
data set to construct highly accurate conformational distributions for protein backbone
(Ramachandran plots) and sidechains (rotamers). These distributions are the most accurate
and restrictive produced to date, with improved discrimination between rare-but-real
conformations and artifactual ones.
Second, I analyzed hundreds of alternate conformations in atomic resolution crystal
structures, and discovered that dramatic conformational change in a protein sidechain
is often coupled to a subtle but very common mode of conformational change in its
backbone -- the backrub motion. Examination of other biophysical data further supports
the ubiquity of this motion.
Third, I applied a model of backrub motion to protein design calculations. Although
experimental characterization of the designs showed them to be unstable and/or inactive,
the computational results proved to be very sensitive to changes in the backbone.
Finally, I describe how MolProbity uses my conformational distributions together with
all-atom contacts and other tools to validate protein structures, and how those quality
metrics can be combined visually or analytically to provide "multi-criterion" validation