parsing pdb for support vector machine

Question

I am working on SVM (support vector machine) using python.

My input is pdb file. I have seen number of examples in which dataset is taken. But, in my case, I want to parse 3d structure of protien (.pdb file).

How Can I get the svm from pdb?

Answer 1

SVMa are designed to work with numbers. Proteins 3d structures are far from being such. There are few possible options:

• Create some kind of fixed length, numerical representation, this is called fingerpint in cheminformatics, and includes dozens of existing methods ie. KlekotaFP; PubChemFP, subFP etc.
• Use a custrom kernel, developed for such structures (like labeled graphs), one such kernels are graph kernels for cheminformatics, ie. those included in chemcpp