From my experience protein-based classifiers like Kaiju and CAT are pretty good at detecting prokaryotic vs eukaryotic contigs. In addition, proks contigs are almost always classified (even when only 1kbp long), while euk ones often not

Does the sense of cringiness when doing that in fromt of another person decrease?

Never used for isolates, but yes, I also agree with your assumption

In case of metagenomes, phyloflash (hrgv.github.io/phyloFlash/) is an option

This might be very basic, but what about generating different scenarios with different hypothesis (e.g., different eqs to describe a given phenomenon) and focusing on comparing the impact of those hypothesis?
Could also be couple with sensitivity and uncertainty analysis for various parameters

Do you loop over the samples or parallelize them? I usually run tasksblike this in parallel using a Slurm array

Last but not least, I want to thank @drinkmic.bsky.social
and @alecava.bsky.social for all the discussions and support through this journey into the Legionellaceae family!

n/n