Metagenomic Assembly and Kraken2

Question

[gtasource]

Hi all,

We have some metagenomic shotgunning data that we assembled into contigs using SOAPdenovo. Now, we wanted to feed these contigs into Kraken2. However, I've looked over some literature, and it seems like most people feed the raw reads into Kraken2. Is there any advantage using contigs? Or because of the k-mer approach that Kraken2 uses, we are actually doing ourselves a disservice.

Secondly, if we don't use Kraken2 with our contigs, what are some options we can use with our assembled contigs?

[Asaf]

You can definitely use kraken2 for contigs. Other tools to determine taxonomy of contigs (or better contig bins) are gtdb-tk, checkM and many more.

What should you do with the contigs? Well, I hope you didn't spend sequencing money without a purpose.

[cschu181]

Well, I hope you didn't spend sequencing money without a purpose.

Isn't that how sequencing money is usually spent? >:D

Answer 1

[h.mon]

In case you are afraid Kraken2 kmer approach will yield false positives, adjust the --confidence parameter - the default is zero, and I feel this is not appropriate for metagenomic contigs.

A small increase will diminish the number of false positives, but my experience is this parameter is really sensitive: at 0.2 one will already achieve a large reduction in classification, and something like 0.5 will reduce classification to near zero - these figures also depend on how well the database represents the composition of your samples.

  --confidence FLOAT      Confidence score threshold (default: 0.0); must be
                          in [0, 1].