Add a manual template?

[rafwiewiora]

I want to add a manual template to the modeling (unpublished structure). @sonyahanson John said you've done this before? Do you have any quick pointers / script?

Thanks so much!

[rafwiewiora]

Ok, so this is pretty simples, yet hacky unfortunately. All you need is, after ensembler gather_templates and before ensembler align:

• into templates/structures-resolved provide a cleaned up pdb
• append a FASTA sequence of the PDB's topology to 'templates/templates-resolved-seq.fa'

So for now, I input the manual PDB in a manual_pdbs directory and run this between gather_templates and align:

import mdtraj as md

# put manual pdbs to be added in manual_pdbs/
manual_pdbs = ['TDIX.pdb']

for pdb in manual_pdbs:
    traj = md.load('manual_pdbs/' + pdb)
    protein_atoms = traj.top.select('protein')
    traj = traj.atom_slice(protein_atoms)
    traj.save('templates/structures-resolved/SETD8_HUMAN_%s_A.pdb' % pdb.split('.')[0])
    resolved_seq = traj.top.to_fasta()[0]
    f = open('templates/templates-resolved-seq.fa', 'a')
    f.write('\n>SETD8_HUMAN_%s_A\n' % pdb.split('.')[0])
    f.write(resolved_seq)
    f.write('\n')
    f.close()

The only obstacle I had for just passing the PDB to Ensembler API was the use of SIFTS files - will need to write something extracting the appropriate features from the PDB. Let me think about this a bit more and propose something (ultimately I think having to call gather_templates twice would be good - e.g. first gather from Uniprot, then ensembler gather_templates --gather_from manual_pdb --structure_path X).

@danielparton do you have any suggestions? I was wondering a couple of things about the code:

• what's the significance of the templates-full-seq.fa file? Here I'm skipping writing the full sequence of my manual PDB there with no problems
• you remove some residues before making the FASTA for templates-resolved-seq.fa - particularly mutations - why is this important? Here I've just pasted the FASTA for my manual structure 'as is' with no problems - in fact I preferred this way, because when I removed mutated residues, ensembler align was generated an alignment with one residue aligned wrongly (it was an R surrounded by mutations, and there was another R just before those mutations removed - so it was clicking as that first R not the right, second one).

[sonyahanson]

Sorry for the delay here. I like your script for adding the new templates to the fasta file programmatically. I have just done this manually so far. I have a pretty complete description of how I've been using ensembler in the astrazeneca and dansu-dansu repos.

[rafwiewiora]

Cool, thanks!