I am using Modeller to construct a homology model of my query sequence (250AAs) from two template proteins (A: major protein of 225 AAs and B: minor protein of 25 AAs). The overall identity of the query and template is very high (~85%). There is a loop of 20-25 amino acid, which is constructed using the template B. To my surprise, in the final model, the constructed loop of about 20 amino acid is very weird and is folding back into the protein (sandwiched between secondary structure) resulting in numerous clashes. I feel this is unusual. Could anyone suggest me the reason for this and the probable solution as well?
*---*Thanks & Warm Regards*---*
Malkeet S. Bahia, PhD Post Doctoral Fellow at the lab of Prof. Hanoch Senderowitz PBC Indo-Israel Post-doc fellowship Dept. of Chemistry, Building 211, Room no. 211 Bar Ilan University, Ramat Gan, 5290002, ISRAEL
Mobile: +91-9815988501 (India, WhatsApp only) +972-584553434 (Israel number)
On 4/7/19 6:36 AM, malkeet singh Bahia wrote: > To my surprise, in the final model, the constructed loop of about 20 > amino acid is very weird and is folding back into the protein > (sandwiched between secondary structure) resulting in numerous clashes.
20 residues is a very long loop - you will need to build a *lot* of models (>1000) to stand a chance of reasonably sampling the conformational space. Once you filter out the poorly-scoring models, you should be left only with non-weird loops.
In some cases, the starting model (.IL file in PDB format) may be generated such that the optimizer can't recover. In this case you may want to look into overriding the build_ini_loop() method: https://salilab.org/modeller/9.21/manual/node95.html
Ben Webb, Modeller Caretaker