[Discussion] Testing on microeukaryotic organisms as a negative control

[chklovski]

[Discussion] Testing on microeukaryotic organisms as a negative control

[chklovski]

Thanks for noting this! The way CheckM2 was trained was using a ceiling of ~35-40% contamination, meaning that contamination numbers beyond that do not scale meaningfully with real biology. A reported contamination score of more than 35% is CheckM2 stating the genome is of very bad quality and is unusable, regardless of what it is or how complete it is.

If you throw enough genes at it however, completeness scores do seem to creep up. We will look into whether explicitly training CheckM2 to disregard eukaryotic genomes as 0% complete may help.

[jolespin]

What about using the eukaryota_odb10 marker set to make a small false positive model? That is, if it has these then it's likely a eukaryote and flag it as such? I admit I haven't read the methods for your paper yet so this might not be applicable at all.