I am wondering if there is any software that can tell, given an uniprot accession and a site, whether this site falls within some predicted structural feature such as alpha helix, beta sheet, idr, etc. I have a list of PTM sites in several hundreds of proteins, and I want to provided statistics of how many of these sites occur in structured regions of different kinds or disordered regions. I know that uniprot, etc., have domain information with coordinates, but I haven't found a good way to output this data for bulk entries in an easily parsable output. Thanks
2 answers
If you are willing do do a bit of work, this is doable. One way of doing it is to set up protein language models:
https://github.com/sacdallago/bio_embeddings
They describe how to make structural predictions for any given protein.
If you prefer a web interface and don't mind doing copy/paste for all your proteins, it can be done by entering either a sequence or a UniProt accession number:
You can use iCn3D in batch mode. Some example Node.is examples are at https://github.com/ncbi/icn3d/tree/master/icn3dnode.
The first three examples in iCn3D gallery show the annotations for AlphaFold structures: https://www.ncbi.nlm.nih.gov/Structure/icn3d/icn3d.html#gallery.
Anything available interactively can be retrieved in iCn3D Node.is script. You can open an issue in GitHub to request a Node.is example for your case.
Log in to answer this question.