Better prediction on genome assemblies than reads about kraken

[tseemann]

288681 is strain specific: Bacillus cereus E33L
86661 is more general Bacillus cereus group

So there was probably statistically enough E33L specific kmers for it to be binned as 33L.

[jenniferlu717]

This has to do with Kraken's base algorithm. Kraken tries to give the most specific classification possible.

So for example, if two species A/B are in genus X and 2000 kmers classify at genus X but 30 kmers classify at species A and 2 kmers at species B, Kraken will give a classification of species A.

More generally, Kraken classifies reads at the leaf level of the root-to-leaf path with the most kmers.