Hello there,
I used the LOMETS server for prediction of a recombinant protein (180 AA) attached with a 20 AA-long peptide tag, of which the secondary structure is unknown.
The mature protein structure of the 180 AA-protein is published and served as template. The predicted structure was matching quite well my expectations: The mature region was overlapping with the template, as expected, and the attached 20 AA-long protein tag was exposed outwards.
I was wondering, how the prediction was done. I guess the mature region was modelled by protein threading, right? Was the unknown peptide sequence modelled by ab initio folding? And were both prediction combined afterwards?
Which method is usually used for ab initio folding and does it include total energy minimization?
And thank you for the development of the great tools!
Best regards,
Fab
LOMETS
Hello,
You are correct in that LOMETS builds models starting from the threading alignments, where the final structure of the aligned and unaligned regions are built using MODELLER for homology modeling targets or gradient descent-based minimization for ab initio modeling targets guided by the templates and deep learning distance restraints. The method does use total energy minimization.
Dimerization
Thank you for the helpful information.
The model structure received from LOMETS is a monomer, which is actually composed as a homo-dimer. Would it be advisable to create a dimer, composed of the model structure, and to apply the pdb-file (of the dimer) to ModRefiner?