Protein-peptide interaction domains, such as PDZ domains,facilitate the localization and regulation of many cellular processes.Determination of the specificity of those interactions, as well as thestructural features that promote such specificity, is important for broadeningour understanding of molecular recognition. Here we present resultsdemonstrating progress towards computational, structure-basted prediction ofPDZ domain specificity. Our study leverages the large amount of PDZ phagedisplay data that has recently been produced. The first stage of our modelingapproach involves generation of an ensemble of near-native backboneconformations using the Backrub flexible backbone sampling method. The second stagescores approximately 106 sequences generated using a geneticalgorithm. For human PDZ domains, we are able to predict the specificityprofiles at most of the non-flat positions. In many cases, our method correctlyselects the dominant amino acid observed in phage display over the amino acidused during structure determination.

Prediction and structural rationalization of PDZ-peptide
specificity using flexible backbone modeling

239th National Meeting (2010)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.