consistency checks

[amelie-stein]

What consistency checks are currently implemented? And to what extent are they automatically controlled, vs. which need a human to look at, and who does that? Would we spot / be notified if a relaxed structure had >0 REU, or if there suddenly was a large fraction of high-scoring ddGs?

[j0kaso]

I added 3 sets of tests (https://github.com/KULL-Centre/PRISM/tree/dev_v0.1/software/rosetta_ddG_pipeline/test):

• create mode checked by unittest as those should provide the exact same files (testing 4 different inputs)
• mp & sp "scientific" tests: these check if the ddG values can be recalculated and raise an error if corr < 0.85. The values of each run are stored in a comparison file (e.g. for soluble example. For the mp test, those values are also compared to exp ddGs.

The test for the soluble protein (sp) could be expanded.
So far, relaxes or their structures are not tested. Would you @amelie-stein have an idea how or what to test?

[amelie-stein]

Hm, the tests I was thinking of (above) are for individual runs, as each protein and structure are different. They were to set conditions in which a warning should be issued that a structure may not be a good starting point for ddG calculations (high energy, clashes). The tests you list are for reproducibility, which is a separate issue. I don't think we need to test plain Rosetta relax for that, as that is done elsewhere.

[j0kaso]

Ah, got it! Indeed, that something different and I agree, a quality check of a structure would be good to have! Currently, nothing what you mention is implemented

[j0kaso]

Regarding the original question: We do not automatically check for large fraction of high-scoring ddGs as this should be noted by the user and can be due to several reasons. Also I would not know how to set such values.

With respect to the Eres of the relaxed structure, this can be calculated using ../software/scripts/pdb_to_prism - rosetta_energy_to_prism function with the input parameters: relaxed_pdb_file (output from the pipeline), temp_dir (for calculations and information for the prism parser: pdbID, pdbchain, uniprot_id; and outputs: pdb_prism_file
It extracts the energies from the final output pdb of the relaxation (summed up as comments at the bottom of the file).

I noted both things in the README and would therefore close this issue if that is sufficient @amelie-stein.

[j0kaso]

I added now the automatic calculation of the E_res file (a275fe5). I am not sure if we want a checking of large fractions of ddGs but if so, please elaborate about the criteria @amelie-stein

[amelie-stein]

Thanks. The criteria are indeed a bit tricky and might need to be adjusted per project. By default, I’d say 15 ddGs above 20 indicate a problem and are not quantitatively informative. It’d need to be adjustable though. Also not urgent at all, so long users are aware of this issue.

On Fri, 11 Mar 2022 at 17.16, Johanna K. S. Tiemann < ***@***.***> wrote: I added now the automatic calculation of the E_res file (a275fe5 <a275fe5>). I am not sure if we want a checking of large fractions of ddGs but if so, please elaborate about the criteria @amelie-stein <https://github.com/amelie-stein> — Reply to this email directly, view it on GitHub <#17 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA4IWXZGLSCAUDOZCY2Z6NLU7NWUXANCNFSM4OMF7HKQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>. You are receiving this because you were mentioned.Message ID: ***@***.***>

-- Sent from my iPhone