Hi, I didn't yet get across to reproduce, but there was a recent report in #38
that mass rounding might be different between CLI and web version. Could that
be the case here as well ? Yours, Steffen

I don't think it's a mass rounding problem since mass values of pyrimidine and pyrazine are identical.
In short,
MetFrag CLI (for my feature) -> pyrimidine
MetFrag Web (for the same feature) -> pyrimidine and pyrazine
I think MetFrag Web is correct. I want to know why CLI did not find pyrazine.
Thanks.

This could be due to you using a different version of HMDB … the exact database used in MetFragWeb (in the screenshot you show) is available from here: https://zenodo.org/record/3375500 and should be used via the localCSV option with the CLI (rather than an SDF downloaded from HMDB). Can you confirm that this file gives you the same candidates as your web query?

With "localCSV" option and the above CSV file, I could have two candidates as the same as web query. By the way, when I tried "localSDF" and SDF file (same mass query), I got some errors and could not obtaine the result.
It would be great if I could use a SDF file directly. Most databases support SDF file, but not CSV. The converter (SDF to CSV) did not work well for the latest HMDB SDF file.
Thanks.

We also had issues with the HMDB SDF, hence we created and validated the CSV file. It's easier to add extra functionality to the CSV than the SDF. However, these local CSVs are a temporary solution and will eventually be integrated into a database so that you do not have to specify external files. This functionality will be released when ready but is not yet available.
Please use the CSV for now, note we have a much larger range of CSVs available at the moment than SDFs, but if you want a flexible workflow, maybe it's best to work with both formats.