Hi Lorenzo, If the structure for your receptor is unknown, then you can use Homology Modeling methods to get a rough idea of the structure, MODELLER is a well know tool for this (http://salilab.org/modeller/). Of course depending on your % similarity to the template, the higher the % similarity, the more reliable your structure may be (of course assuming there are no major conformational changes, etc.)
Now, to figure out the sites of interaction, you could use a shape based complementarity approach like the one used in the ZDOCK algorithm (http://zdock.umassmed.edu/software/). This gets to be a little bit trickier if your % similarity to your template is low, because the dissimilarity is often due to surface residue differences, which are obviously the ones you're interested on. On the other hand, if the source of interaction is driven mainly by hydrophobic forces, then an analysis using the spatial aggregation propensity method (http://pubs.acs.org/doi/abs/10.1021/jp911706q?journalCode=jpcbfk) may reveal interesting sites of aggregation. This method is a little bit more forgiving that the shape complementarity one because of the intrinsic averaging that goes on to determine the site of aggregation. All of these methods and other simulations tools are available in the Discovery Studio suite from Accelrys. Disclaimer: I work for Accelrys as their Product Manager for the Life Science Modeling and Simulations suite of products. So, if you're interested in evaluating and gain access to these tools please contact me directly. Kind regards, Francisco Sr. Product Manager http://accelrys.com From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Dr. Lorenzo Finci Sent: Thursday, August 02, 2012 6:07 AM To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Protein-Protein Interactions Dear Colleagues, I have a question for all of you bioinformatics oriented structural biologists: How do I predict the sites of protein-protein interactions between two receptors that have been proven to interact biochemically but lack specific details regarding proximity. This is not a straightforward question for me, and I believe it is somewhat complicated. The complicated scenario involves a multitude of different subunits and isoforms. Also, there is not structural data to support all components involved, and thus I presume I should use the sequence based software. I am aware that there are different types of prediction software, either sequence or structure based predictions using different algorithms: http://rosettadesigngroup.com/blog/58/10-protein-protein-interface-prediction-servers/ Receptor 1: -Has 5 predicted subunits (Alpha)2-(Beta)2-(Gamma)1 1. Alpha (6 isoforms) 2. Beta (3 Isoforms) 3. Gamma (3 Isoforms) Receptor 2: -Is believed to be composed of (Alpha)3-(Beta)2 1. Alpha (4 isoforms) 2. Beta(1 isoform) Any advice or recommendation will be well appreciated! Sincerely, lorenzo Lorenzo Ihsan FInci, Ph.D. Postdoctoral Scientist, Wang Laboratory Harvard Medical School Dana-Farber Cancer Institute Boston, MA Peking University The College of Life Sciences Beijing, China
