12.07.2015 Views

Protein Engineering Protocols - Mycobacteriology research center

Protein Engineering Protocols - Mycobacteriology research center

Protein Engineering Protocols - Mycobacteriology research center

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Combinatorial <strong>Protein</strong> Design Strategies 17Fig. 3. Probability distributions of amino acids (upper) and nucleotides (lower) forsite 54 of a SH3 domain (PDB:1CKA). For the amino acid probabilities, the desiredprobability distribution is shown by open bars, and that encoded by calculated genelibrary is shown by filled bars. An oligonucleotide library with the frequencies of thenucleotides specified in the lower panel encodes for the site-specific amino acid probabilitiesin the upper panel.the computed nucleotide probabilities (shown in the lower panel of Fig. 3). Theagreement between the two is excellent in this case. In general, the calculatedprobabilities agree well with the desired probabilities. In many cases, an exactmatch between the desired and calculated probability distribution cannot beachieved because of the partial degeneracy of codons to amino acids. This computationalmethod provides excellent yields of complete sequences (those notcontaining stop codons): for test proteins of 50 to 60 residues, in which all sitesare subject to selective randomization, the yield of complete sequences is 96%or more. High yields are particularly important when a large fraction (or all) ofa gene is subject to combinatorial mutation.3. Notes1. Self-consistent methods are not the best for finding global optima (16). In proteindesign, this is generally because of the approximate manner for treating correlationsbetween residue identities and conformational states in mean-field-like theories.Stochastic sampling and elimination methods generally provide better optimizationresults.2. In designing sequences, backbone flexibility should be considered. In natural proteins,small backbone adjustments can accommodate mutations (70,71). However,most computational methods treat the backbone as rigid because including theflexibility is computationally demanding. The probabilistic nature of the statisticaltheory suggests that its predicted profiles are less sensitive to backbone choice.Amino acid profiles from 21 slightly different backbone structures of protein L aresimilar to one another for energies (or effective temperatures) above the peak of

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!