Science Buddies: "Ask an Expert"

Hi all,

I'm trying to implement an empirical scoring function by using a training set of RNA quadruplexes. Given a bunch of PDB files containing ligand-bound complexes, how do I perform linear regression on these files to obtain the scoring function?

Having a hard time finding a general 'tutorial' for this kind of thing.

Thanks!

- Eric

ericjang,

It depends on what your criteria are. What will you consider "good" and what would be "bad?" By PDB I assume you mean Protein Data Bank, so I further assume that you will be considering some aspect(s) of the molecular complexes as scorable, such as conformational energies, hydrogen bond distances, bond angles, etc.

What are you examining in the PDB structures? What are you doing this for, are you evaluating some kind of structural variation?

Hi hhemken,

I hadn't thought of it that way... basically I am trying to do a virtual screening project. I know that there are a lot of software packages out there to do this but I wanted to learn this for myself down to the implementation level, among other aspects of my project.

I think I need to back off and ask a more 'general' question in another thread so I can understand what concepts pertain to my project. I suspect I might not even be able to use empirical scoring because there are few structures (pdb code 3IBK) actually available with bound complexes. Most likely going to be force fields, which I have a gazillion questions about.

Thanks for your patience, it really means a lot to me

- eric