Maintain residue numbers from a pre-build model & phosphoresidue modification

[TueRens]

First of all, thanks so much for building the software, it is truly excellent.

I have two problems I run into where I could use your help, if this is even possible:

1. If you start from a pre-build PDB-file where only a couple of residues are missing, is it possible to maintain the residue numbering from the original PDB file?
2. I have a complex with a protein (all normal AAs), a peptide (containing normal AAs and one phosphorylated residue) and a ligand (three-letter code XXX). For some reason, if I use the contents.json file below (I changed the protein sequences), it does build the first protein just normally. However, it doesn't recognize the second protein (which is the peptide) and builds all alanine residues for this stretch, labeling them as UNK. In the pre-build model, the phosphoresidue, along with some normal residues flanking it, is already build in. Strangely, when I use the FASTA approach, all is fine. However, I don't think there is a way to apply the modification in that case. I'm not even sure if the phosphorylation is possible using modelcraft. I prefer to use the json approach, since it gives more control. Any advise?

Here is the representative contents.json:

`{
    "copies": 1,
    "proteins": [
        {
            "sequence": "THISISARANDOMSEQUENCETHATIJUSTCAMEUPWITH",
            "stoichiometry": 1,
            "modifications": [],
        },
        {
            "sequence": "ANDSOHEREISANOTHERONE",
            "stoichiometry": 1,
            "modifications": ["4->SEP"],
        }
    ],
    "rnas": [],
    "dnas": [],
    "carbs": [],
    "ligands": [
        {
            "code": "XXX",
            "stoichiometry": 1
        }
    ],
    "buffers": [
        "NA",
        "MG",
        "CL"
    ],
    "smiles": {}
}`

[paulsbond]

Thanks for getting in touch.

Regarding the first issue, this is not currently possible. The protein models are passed through Buccaneer and this reassigns the sequence. All being well, the output sequence will be correct and the numbering will start from 1 as the first residue of the sequence. I agree it would be good to match the numbering to the input model in the case where the input model is very complete.

The second issue is related. The JSON format is useful as you can specify the number of copies (only used for a solvent content calculation for density modification) and the sequence types in case they are ambiguous. The modifications are only used to tell Buccaneer to build selenomethionine (M->MSE) at the moment. Other modifications won't affect the building at all, although the format was created with automated modification building in mind for the future.

The input sequence and side chains are not used and Buccaneer may not be able to able to guess the correct sequence for short chains. Small changes such as one extra residue or small differences in the backbone conformation can affect whether the sequence is guessed correctly or not. Any unsequenced residues are just given the label UNK.

I hope to make ModelCraft better for model completion tasks such as this in the future. At the moment it is designed to get you from a partial model to a mostly complete model, and interactive validation and completion in a program such as Coot is still very much needed.

[TueRens]

Thanks for your detailed response. Thanks for confirming that there is currently no good way to transfer the residue numbers. I'm working with a protein where we see some scarring as a result of cleavage (leading to minus-residues) as well as a peptide fragment from a protein. So, this makes it a bit challenging.

I indeed did some testing with with other models, and there Buccaneer was able to pick up the residues. I do have to play around with things a bit more.

From my initial testing, ModelCraft can indeed be very powerful for model completion. I'm working with protein-peptide-compound complexes where the compound induces significant changes in the peptide conformation. I was very excited to see that ModelCraft was able to accurately build these significantly different conformations in a very good way. In some cases even better than what some of our experienced crystallographers were able to build. The iterations take quite some time if you are only interested in the changed conformation of the peptide chain, but I envisioned it to run these tasks in parallel overnight, or in the background, since we are dealing with >200 crystal structures each month.

I will play around with it a little more, since it has so much potential. But, as you understand, I would be very happy to see this model completion module pop up as an update somewhere in the future!

Thanks again.