merkys / covid-lt Goto Github PK

breaks snakemake -S

voronota related rules should use a docker imgae with installed voronota. All rules with external dependancies unavailable with conda should be run under a docker image.

• code/workflow files could be separated from temporary files or results foldes. In config file could define lets say a "output" directory and use it for all folders that contains files created by the workflow. This would be good to run a separate small testing run
• sort of integration test is needed. I would suggest to use a small subset of initial list of all files (let say 10 structure listed in the configs/defaults.yaml)
• if it would not take too much time its worth considering rewriting the rule "quality_map" in a moree typical way for snakemake. Ideally there should never be any for loops in the shell part and multiple
  also a script bin/pdb_renumber_S1 culd be run in a separate rule.
• why there is 1500 in the qualit_map rule " join -a 1 <(seq 1 1500 | ", could it cause a crash under some circumstances.
• I guess in places with "pdbid=downloaded_pdb_files()" it would be good to use checkpoint approch as we dpn't know the exact number of PDBS to be downloaded.
• in some rules like propka_out , temporary directory is created in the same folder as Snakefile. Its a bit risky - it would be good as discussed above to provide a "working" directory via a config file and provide this to the rule via "params:" keyword
• an output file if possible should not mark a completion like "download_pdb_all.log". In this case an output could be a folder with downloaded pdbs. ( please look at directory() )
• ideally commands should be runnable from the coda environment defined in envs/covid-lt.yaml or from rule specific conda environments or docker images (if conda environment do not work)
• what is "..tab"
• from file alignments should be split into folders containing created/modified files by the workflow and reference data

It is expected that the PDB outputs in pdb/Chothia/ have their antibody chains renumbered according to Chothia scheme. However, antibody_numbering_converter does not seem to do that properly. For example, chain E of PDB entry 7WD9 would not contain two residues indexed as 52: 52 and 52A although this is what AbNum would output.

Edit: Even worse situation observed with 7K8X chain D which in Chothia's scheme has multiple residues numbered 100 (with letters A-M) and antibody_numbering_converter ignores that completely. Thus the question is whether this tool is needed at all. An alternative tool is ANARCI which is a Python package and could be used locally. However, question persists whether the PDB format will allow multiple residues of the same number.

No binary or dockore for antibody_numbering_converter - somehow the workflow should be transportable

In branch blast-promod-split-chains, commit 1e6885a script bin/contact-graph fails to correctly identify S1 chain in PDB complex 7RXD. This is due to the fact that H and L antibody chains are also contacting with maltodextrin-binding protein. This chain is detected as S1.

It would be nice to agree upon a license for the code in this repository. I saw binding_evaluator has Apache 2.0. Maybe we could use it for this project as well?

:) never ever touch output files like " bin/voronota-contacts -i {input} > {output} 2> {log} || touch {output}""

Hi,
I tried to make up a small testing set - please look at the gaplay files.
I used the conda environment defined in envs/covid-lt.yaml.
After running snakemake -f all -c 8

I get

But outout files are ampty

• i think we could start using configs for specification of output files and etc.
• some parsed PDB databases could be saved as LFS

rule profix on the error outputs a blank pdb file and no error report.
Maybe rule profix could be divided into separete tasks for clearer debuging. I would suggest using snakemake log files and keep "bash" part as minimal as possible.