I’ve been wanting to make a statement about contaminants, something like “These pairs of contaminants occur together more than would be expected by chance.” But i was recently reminded me that if two organisms co-occur at all, it’s going to be more than expected by chance just by occurring at all, given the potential number of pairs.
A work around i thought of this morning would be to:
assign a species or genome a biome score based on how frequently that species occurs in a given biome, or how general or specific that species is across biomes. I would use the ENA biomes, and I think there are only a few thousand (30k?) ENA metagenomes. I’m pretty sure running full gather with gtdb rs207 on all of the ENA metagenomes is not really a big deal computationally.
with the biome score, I could do statistics on whether contaminants are more likely to be from the same biome as their host, or not. (if they are, then those contaminants were probably present in the “isolate” culture, and if they’re not, then those contaminants might be from technical sources like extraction kit or lane hopping).
I’ve been wanting to make a statement about contaminants, something like “These pairs of contaminants occur together more than would be expected by chance.” But i was recently reminded me that if two organisms co-occur at all, it’s going to be more than expected by chance just by occurring at all, given the potential number of pairs. A work around i thought of this morning would be to: