I am developing a program for ab initio protein folding using the genetic algorithm. I am creating sets of phi and psi values at random for the required population (as is the first step for GA) but I only wish to make this randomization within allowable ranges. Hence I need a database of all allowable phi psi POINTS with the values. Everywhere I check I find only graphs or the values of the boundaries.
I expect something like:
phi psi
x1 y1
x2 y2
x3 y3
. .
. .
By one click you can save Ramachandran plot values by using SPDBV.
File >> Select Your PDB file>> Open
Select >> All
Window >> Ramachandran plot (Optional for visual inspection)
File >> Save >> Save Ramachandran Plot Values.
Thats All.
Column Order in saved will be Omega, Phi, Psi
This will also enlist another column of secondary structure before Omega Column in saved file.
'Allowed' Ramachandran values follow probability distributions. The occupy-able values are also not discreet, they are continuous. So two suggestions 1) easy way... Pick some cutoff for minimum acceptable probability and then define bins (1 degree, 0.1 degree, 0.01 degree?) within those limits or, 2) Sample from the distribution to get a more realistic representation of the allowable (and diversity of) possible conformations.
Here is a site from which you can get nearest neighbor depended Ramachandran probability distributions....
http://dunbrack.fccc.edu/ndrd/
I did not find any data in that website. However I did find a site where they gave a list of 500 pdb files which they used to generate a distribution of torsional angles in the phi psi phase space. It is an approximate approach. They are real proteins and therefore will give a general view of the allowed phi psi values. I generated a database of all phi psi values from these proteins (say if each is 100 residue long then 500 proteins gives 50000 phi and psi values) and rounded them off to the first decimal.
Do you think this database is worthwhile ?
Dear Archisman,
Your effort is valiant but somewhat misplaced. In order to fold proteins we all need to first understand the principles of coexistence in complex systems. In another words the physics that do not exist yet. Regular energy based minimization cannot possibly work. As the works of the last 5 years show so clearly proteins are a composite of solid-like and liquid-like components. These are not well defined and can interconvert under environmental conditions (context). Therefore usage of genetic algorithms which code for set of rules is just a misguided effort. Neglecting the fact that a bunch of other people tried this effort proteins are always ensembles of differently folded individual molecules. If you can somehow produce the part of ensemble that would be of value but I do not know how to code for that as nobody really produced any real ensembles. My modest attempt described in (Burra et al PNAS 2009) was very limited to the PDB structures (that nota bene you were planing to use). Molecular Dynamics is a very limited method to provide insights into the real ensembles.
Coming back to your direct question how to get the sampling for the Ramachandran representation you should contact Rampage authors. This is a relatively well composed representation of the protein backbone allowed states even though a real quantum mechanics based representation would be better. Good Luck with your attempts.
Sounds like a good start. The identity of the two resides matters a lot... meaning the confrontational distribution of a Gly-Gly bond is not the same as a Gly-anything else... The data on that site is not directly available for download, you have to request it under an academic license.
You can find allowed protein backbone torsional angle (phi and psi) values here:
http://kpwu.wordpress.com/2011/04/19/making-ramachandran-plot-ii/
You might find this helpful
http://onlinelibrary.wiley.com/doi/10.1002/prot.10286/abstract
- Badges
- Science topic
- Similar topics
- Papers