is there a 3D protein structure format on amino acid level?

Question

[Yannick]

I need the protein structure to predict possible phosphorylation sites. For this i want to use the 3D structure data. But structure data on atom level is quite large and i'm not sure that the neural network will deal well with just plain atom coordinates. That's why i wondered if there would be different encodings of protein structures such as index, name, point position, rotation This would offer a much more compact format of 3D structure, decrease the feature dimensions, smaller filesize. I already know of .pdb and .mmCIF that are available as downloads on pdb

Answer 1

[Mensur Dlakic]

You may want to look at FEATURE, which extracts local descriptors from PDB structures:

https://simtk.org/projects/feature

Answer 2

[Jiyao Wang]

You could just use the C-alpha (atom name “CA”) atoms to represent each residue if necessary.