Hi @RJMW

To have the tools from https://github.com/computational-metabolomics deployed more easily it would be cool to have them all in one repository. This would simplify maintenance a bit:

• common github CI
• we could add them to https://github.com/galaxyproject/planemo-monitor (which would automatically create biocontainers for tools with more than one requirement)
• we could add them to https://github.com/planemo-autoupdate/autoupdate (which creates automatic updates if there are new conda packages)

If you create a new empty repo I would volunteer to make a start for some of the tools. Issues could also be moved.

Alternatively, we could also move the tools to one of the larger tool repos (e.g. IUC or W4M). There you only need to grant write access for the respective organization in the toolshed repos.

Training data should go to tool-data (by using a tool data table). Then this is configurable by the admin. Maybe we could allow to select different ones or from history. Even if this might not be used at the Moment the would help for the future if model generation becomes easier, eg if the corresponding galaxy wrapper is available.

I was wondering if anybody knows if it is possible to output an InChiKey?

The example outputs at the moment include an InChi structure format but not an InChiKey. We could potentially do the conversions after but if it is already created somewhere it would be useful to include in the output

I would suggest to accept msp files with space separated columns.

Applies to metfrag as well.

This would immediately solve #4 and would remove a lot of maintenance burden on making this tool usable.

It would be an option to have the data in a Galaxt data table (.loc file etc) directory. In particular if there are multiple possible models available. Then we could also remove the reference to __tool_data_directory__

Also allow users to update their own models could be nice.

I think that a for details of the python script should be documented in the help section: eg

You should check the return code of this

Does one call of the python script maybe generate multiple calls to cfm-id. I'm just guessing here. Could allow for some paralellization, eg on Galaxy side.

The local database in cfm-id and metfrag seems very similar. But its csv in one and tsv in the other. Maybe it would be good to have this consistent. I would prefer tsv since InChI contains commas (question would be if this works anyway with csv if there are additional commas). Then the same input can easily be used in both files (ie the metfrag input can be used in cfm-id - since the first columns are the same)

Maybe we should allow omitting InChI or SMILES (in the python sources there is a TODO anyway). Due to the use of the python wrapper we are free to define the format as we wish. Possibilities:

• Allow to keep one column empty (maybe this is possible anyway)
• Make column 3 inchi/smiles .. let the python check for startswith("InChi")?
• Detect by column header