Question about Kraken2 core_nt vs nt database results.

Question

[Mark]

I'm trying to run Kraken2 on a sequencing reads from a lot of novel, eukaryotic samples. I'm looking to choose an appropriate database for this purpose and the most up-to-date on the Kraken2 databases website (https://benlangmead.github.io/aws-indexes/k2) is core_nt. However upon researching, it seems that the new NCBI core_nt database does not include eukaryotic chromosome assemblies (unless I am mistaken). This may be fine for things like BLAST where typically the annotated genomic regions are of interest, but for a program like Kraken2 that performs a genome-wide k-mer analysis, I'm struggling to understand how the new default database is suitable for this program. Isn't it missing a lot of important information and doesn't this reduce the accuracy of the taxonomic predictions?

Answer 1

[GenoMax]

Data size for nt has become so large that many users don't have the necessary hardware available (storage/memory) to effectively use that database. Even kraken2 authors may be struggling to make these indexes in a timely manner. If you have the necessary resources available you could build your own nt kraken2 database. Program authors are making the pre-made indexes available as a courtesy for users. An end user needs to decide suitability depending on their application.

You may actually want to get RefSeq genomes, Representative Assemblies (https://www.ncbi.nlm.nih.gov/datasets/genome/ ) and create kraken2 databases, if you want to maximize genomic content.