Re: [modeller_usage] Loop Modeling

[Thomas Evangelidis <>]

On 12 January 2011 06:04, bharat lal <">> wrote:

Hi,

I have a question regarding modeling the loop of an already determined structure of a protein de novo. I want to change the entire sequence of the loop regions of the protein with random sequences and also I want to extend the length of the loops in the protein structure. I have searched through a lot of papers but I am not able to find any specific paper about such an approach.. So, I want to know can Modeller help in solving this problem and also I will be highly obliged if anybody can refer me to some good paper on De novo loop modeling ..

1. Remove the loop region form your template.
2. Align the resulting template sequence with the same sequence but include the desired aa sequence. The alignment shoud look like this:

>template
AGF--------GHK
>target
AGFRRKTYDERLGHK

where RRKTYDERL is the sequence of the loop you want to model.

3. The use loopmodel build a new model and the loop conformation ab initio as in this example:

http://www.salilab.org/modeller/manual/node32.html

homology models will be named as "target.B9999*.pdb" whereas loop models "target.BL999*.pdb"

4. Create as much as possible loopmodels by setting the following parameters accordingly:

a.loop.starting_model
a.loop.ending_model

Also bear in mind that loop modeling works well for loop <12 aa. When the length exceed ~15 aa the reliability drops abruptly.
I would recommend searching the mailing list about loop modeling, this subject has been covered extensively in the past.


Thomas

PS: QUARK ab initio server has a new option that allows distance restraints. The problem is that it has a 200 aa limit and you have to include the surrounding environment of the loop in order to get an accurate prediction. I intend to try this option soon, so I'll keep you posted.