try using HOMER with conda

[sreichl]

HOMER still gets installed "manually" via a rule.
This occasionally leads to hard to find/understand problems in peak_calling step.
Maybe it's time to try to make it work for conda.

https://anaconda.org/bioconda/homer
https://bioconda.github.io/recipes/homer/README.html
https://github.com/bioconda/bioconda-recipes/tree/master/recipes/homer

[...]homer writes its DBs to: $CONDA_PREFIX/share/homer/data for conda and this approach works.

still requires update / install of new DBs via configureHomer.pl

• put Homer log into the respective log file, as it is not used downstream, shown here
• where is Homer installed if $CONDA_PREFIX is not set by user -> seems to point to .../.snakemake/conda ie the default so can be used
  • set HOMER_path in Snakefile using $CONDA_PREFIX and print as easy test once with global variable and once with login without global variable ie commented in the bash.rc file
• inform rules about homer's successful installation of genomes. how?
  • by checking the existence of the respective genome data$CONDA_PREFIX/share/homer/data/genomes/hg38.tss -> double check if this actually exists.
  • peak_calling
  • homer_region_annotation
• add homer using bioconda to envs/macs_homer.yaml and comment homer dependencies that are manually added (perl, wget)
• update homer installation to only install genome info https://github.com/epigen/atacseq_pipeline/blob/main/workflow/rules/resources.smk

[fwzhao]

One very helpful intermediate step would be to copy/redirect the stdout stderr of peak_calling shell script to a log that is kept. Currently, when peak_calling fails, the respective "fail" logs for each individual step are lost, and it becomes difficult to debug.

Before this issue is closed, I'd suggest a fresh install of HOMER (delete previous path) if anyone has issues that don't make sense. In my case, it was of an outdated HomerConfig.pm file, that existed but couldn't be read somehow by Homer - "Can't locate HomerConfig.pm in... "

And the only thing that fixed it was a complete reinstall (not just perl configureHomer.pl in pre-existing Homer path)

[sreichl]

yes, good point. should redirect all log to {log} directive but without messing with the multiQC setup.

[sreichl]

Add check if $CONDA_PREFIX is set and error of not with message to set it

[sreichl]

idea: consider using Snakemake wrappers
https://snakemake-wrappers.readthedocs.io/en/stable/wrappers/bio/homer.html

[burtonjake]

+1 to this issue. As-is you can't use the homer installed this way for two people to run the same pipeline on the same shared data, since configureHomer.pl hard-codes paths to the installing users (whoever gets to atacseq_pipeline_install_homer first) binaries. Simply trying to reconfigure homer doesn't work because its a permissions mess. The only solution is to force snakemake to install HOMER again, which seems suboptimal.

[sreichl]

@burtonjake Indeed, as you can see, this has been an issue for us for a while, and I would love to address it. It's not that straightforward because it does not seem to work well with vanilla conda, and we've stuck with the solution so far. Any effort in this direction is very welcome (notes for your consideration are above).

Just to clarify: the installation in the resource folder that we share does not suffice. What is the exact problem?