docking with glycan on rosie sever and get this error

[tla268]

I am attempting to dock a glycosylated protein with a standard protein on ROSIE and keep hitting an error. Since I don't have cluster access to run the full glycan protocol, I'd appreciate any help identifying the problem from the attached crash log.

ROSETTA_CRASH.log

[roccomoretti]

So it looks like the initial error is a Segmentation Fault with the backrub step. Segfaults are sometimes hard to debug even when you're running things locally, so I'm not sure I can tell you more with just the server crash output. (That said, segfaults are 100% an issue with Rosetta. Even if you did something silly with your inputs, Rosetta should be catching that and giving you a decent error message, rather than giving a segfault.)

To debug this further, I would need to have access to your input files. (If you don't want to post them here, I should have access to the job page if you post a link to that here, alternatively you can send me them privately by email. The address should be in my Github profile.)

[tla268]

send the private email
Thank you

[roccomoretti]

By the way, you should be able to run the docking protocol even on a local machine (you don't need cluster access). The tradeoff is that it will take much longer with a few CPUs, versus if you can split it across hundreds of CPUs on an HTC cluster.

[roccomoretti]

Thanks for the reproduction case. I managed to track down the issue to some data that was being carried forward between proteins. (In your case, between the receptor and ligand, which causes the segfault due to their different sizes.)

The fix for future Rosetta versions is in PR #758, but that's not going to help with ROSIE (at least not for a while). One option would be to flip the order of chains, such that you put the smaller chain first.

---

That said, after looking into the ROSIE application and your example case more deeply, one issue you're going to run into is that the ROSIE dock application (the https://r2.graylab.jhu.edu/apps/submit/docking one) is built around protein-protein docking. As such, it strips out any HETATM lines (actually, any line which isn't an ATOM line), which means that any of your sugar molecules aren't even going to make it past the input mark. What you'll need to do if you use the server is to change the HETATM lines in the input file into ATOM ones. (This shouldn't affect Rosetta read-in.) When you do, also remove the crystallization adducts, and clean things up to just the residues you're interested in modeling.

This probably won't work though, as usually if you're working with glycans you'll need to provide the -include_sugars option to invoke extra sugar-related processing. Additionally, you'll need -alternate_3_letter_codes pdb_sugar (and likely -write_glycan_pdb_codes) for Rosetta to correctly recognize the PDB codes for the sugars you're using. -- Those options aren't included with the ROSIE application, and there really isn't a way to do so.

Additionally, the connectivity of the glycans is often important, and Rosetta doesn't typically look at the CONECT record in the file for that purpose. It does look at LINK records, and you may need to include those lines for Rosetta to properly represent the covalent connection between the glycans and the protein and between the multiple glycan residues. (Though that won't work for ROSIE Docking, as anything but the ATOM records are being stripped out.) Alternatively, there is a -auto_detect_glycan_connections (use with -include_sugars ) that may do a decent job of things even without the LINK records, but again, that's not available on ROSIE.

If you do want to look at the glycan effects, it looks like you'll have to run docking locally, so you can add the additional flags to properly represent the glycans.