Center for Bioinformatics at KU

As part of our public Seminar Series
http://bioinformatics.ku.edu/seminars

Dec 2, Fri, 2005
10:00am - 11:00am, Regionalist Room, Kansas Union

Dennis Law

Combining Experimental Information and Computational Methods to Produce Testable Models for Protein Complexes

Since many aspects of life are governed by interactions of proteins with other molecules, prediction of these interactions is of paramount importance. In the absense of an experimentally determined structure for proteins interacting with other molecules, researchers may create models for how they interact. Often there is experimental information suggesting the mode of interaction, and methods that utilize this information to create models are needed. As has been shown in CAPRI (Janin et al., 2003), it is possible to create models for complexes whose structures are unknown using only available experimental information. These models are capable of predicting several interface contacts that are present in the crystal structure. We made several predictions in the CAPRI challenge that were able to predict a significant portion of the interface later determined by crystallography. An attempt was made to model the interaction of the regulatory subunit (RI 113-244) with the catalytic subunit of PKA. A comparison to the recently determined crystal structure of the complex between the C-subunit and RI(113-242) (Kim et al., 2005) shows the prediction correctly identified most of the binding site, but did not identify the correct orientation. The main factor contributing to the incorrect prediction was an unexpectedly large conformational change in the R-subunit upon binding (5.8 A RMSD). The conformational change made a rigid body fit of the unbound structure for the R-subunit into the correct position essentially impossible, due to an almost 90 degree shift of the C helix (residues 227-244). By docking the C helix and remaining portion of the R subunit seperately, solutions much closer to native are obtained. This indicates that a fragment based approach may yield better results in cases of large conformational change for one of the proteins. A model for the interaction of RII with the C subunit was also created and the model is consistent with the biological information currently available.