Protein StructurStructural Motifs Guide Predictions for Protein Engineering

energy of all interactions its amino acids make with each other and
with the surrounding solvent. A significant fraction of these
interactions is local, i.e. between amino acids that are close in the
polypeptide sequence. For any given structure of locally interacting
amino acids, there will be a sequence of lowest free energy.
We hypothesize that elements of this sequence can be defined from a
consensus analysis of similar structures. We expect that to a first
approximation the most probable distribution of amino acid frequencies
can be described in formal analogy to Boltzmann's law, which relates
the frequency of observation of a state and its thermodynamic free
energy. Accordingly, we ask how well the effect of a point mutation
on the thermodynamic stability of a structural motif[+] and the free
energy of folding of a protein can be predicted from an analysis of
amino acid frequencies of structurally similar fragments.
To this end, we have compiled amino acid propensities in various
structural motifs from the protein structure database and compared
them with the effects of published mutations in staphylococcal
nuclease, T4 lysozyme, and an immunoglobulin VL domain.

Both the qualitative as well as the quantitative analysis of the
correlation of predictions with experimental data are satisfactory,
if the predictions are restricted to fragments whose interactions
remain in a local context. The qualitative predictions are correct
for 101 out of 141 predictions, i.e. 72 % of the time, for locally
interacting residues. With this method, we provide a tool which
allows the rational prediction of stabilizing mutations with a high
rate of success.