Metagenomics best hit analysis: caveat emptor

I gave a little quote to Ed Yong for his report on the recent study purporting to show the bubonic plague and anthrax on the New York subway.

I thought this would be a good time to put up a slide I use for whole-genome shotgun metagenomics teaching. Namely: what happens if you take an E. coli K-12 reference, shred it into 100 base reads, and then put it through a typical metagenomics pipeline relying on ‘best hit’ analysis:

I think the picture speaks for itself.