Problems using pheval

[seoanezonjic]

Hi Pheval authors
Recently I've seen your paper in Researchgate and I'm very interested so I gave a try. Following the instructions here https://monarch-initiative.github.io/pheval/exomiser_pipeline (I use a previous exomiser version installed in my cluster) I've have the following problems:

• Tutorial saids in "5. Set PhEval Config YAML File" that exomiser directory program key is "exomiser" but it must be "exomiser-USED VERSION". If not set, it breaks with a dict key error.
• The dependency to j2cli to make j2 command works is not defined in poetry files so I have a j2 command not installed error. Maybe the snap version that you use contains jinja2 and j2 cli but I suggest to add "j2cli" to your depencies so all it's handled by python dependency manager
• pheval-utils create-spiked-vcfs gives the following error: "Error: No such option: --template-vcf-path (Possible options: --hg19-template-vcf, --hg38-template-vcf)". Downgrading pheval to 0.3.1 fix this error
• When I execute "make pheval run" the directory configurations/exomiser-13.1.0-default is created but later breaks because the file configurations/exomiser-13.1.0-default/config.yaml not exists. I think that this is counterintuitive, if the folder is created by pheval, I cannot create the config file in a folder that not exists. The same occurs for the preset-exome-analysis.yml file
• The pheval system make symlinks to exomiser configuration files but there is a stage that the file "application.properties" is accessed in write mode. As I work in a cluster and it's no sense install exomiser twice, I referenced the cluster installation that it's write protected so I have a Permission Error. If you simlink this file to write in it (that is not a good practice as the program is also used in other contexts), why not copy it? If you execute exomiser from configurations directory, it will access the copied file instead the original one, so the symlink is not necessary. In fact, I fixed this problem removing the symlink and copying the original "application.properties" file to the exomiser folder within configurations directory.
• Exomiser gives me file errors when with configurations/exomiser-13.1.0-default/2302_hg19/2302_hg19_transcripts_ensembl.ser and configurations/exomiser-13.1.0-default/2302_hg38/2302_hg38_transcripts_ensembl.ser. Pheval do NOT simlink these files or give any error about not finding them when simlinking. Simlink manually 2302_hg19 and 2302_hg38 folders from original exomiser installation fixed the problem.
• File errors with phenotype files. Pheval simlinks each file of phenotype folder instead simlink the folder. Simlink manually the phenotype folder within exomiser folder in configuration directory fixed the problem.
• Due to previous errors I've had several results folders with corrupted exomiser executions. If I execute "make pheval run", make execute and exits from pheval folder instead of resume exomiser execution. To do so, I have to remove manually the results folder. I think that the corresponding rule has to check if exomiser has generated the results files and if not, launch exomiser.

Once I've succesfully executed the pipeline, what is the next step to obtain the performance figures shown in the paper?
Thank you very much in advance
Pedro Seoane

[matentzn]

@seoanezonjic Thank you for your message, the whole team is OOO at the moment, but someone will come back to you when they are back!

[seoanezonjic]

Thank you @matentzn and @yaseminbridges , I''ll wait to the team and in the meantime I'm playing with the platform. In fact, another problem that I faced is that I'm executing in my university cluster but the computing nodes has not access to internet and the script create_noisy_phenopackets.py fails in the following line:
resource = OntologyResource(slug="hp.obo", local=False)
I think that the script should have an option to set "local" attribute to True and used a cached version of the ontology files in this computing environments (and I think that it will speeds up the execution avoiding repeated queries to servers)
Thank you in advance
Pedro Seoane

[yaseminbridges]

Hi, @seoanezonjic!

Thank you for documenting these problems in detail!

I think a lot of these issues may be due to the MakeFile pipeline needing to be updated to what we use in the paper. @souzadevinicius can you confirm?

[souzadevinicius]

Hello, @seoanezonjic,

Thank you for your message. I acknowledge that the PhEval pipeline documentation is outdated. I am currently working on updating it.

In the meantime, you can utilize a pipeline described in this repository: https://github.com/monarch-initiative/monarch_pheval. The repository contains different types of experiments, which can be edited by modifying the PhEval Config YAML File.

To execute all experiments outlined in the config file, please follow two steps provided in the pipeline documentation:

1. Run bash ./resources/generatemakefile.sh
2. Execute make pheval

I apologize for any inconvenience caused by the outdated documentation. If you happen to have any other issues, please feel free to reach out to me.

Thank you.

[julesjacobsen]

@souzadevinicius is the pheval Makefile needed any more if the monarch-pheval repo exists?

[souzadevinicius]

@souzadevinicius is the pheval Makefile needed any more if the monarch-pheval repo exists?

No, it's not, @julesjacobsen. I'm updating the PhEval pipeline documentation and removing these unnecessary files.