Dear authors.

Thanks for your work and these precious comparative results. I'm surprised your fig1.B results. Exact tools, such as Metagraph or Reindeer, should yield the same qualitative results unless the parameters are misconfigured. About bloom-filter based approaches, the best way would be to count the number of unique non-erroneous kmers for each indexed experiment and set the bloom filter sizes depending on those results. I think that you over-dimensioned filter sizes in your experiments.

In case it helps, we just released kmhelpers, which automates the tasks of organizing and sizing bloom filters when building indexes with kmindex. https://github.com/sebllns/kmhelpers/

Best, Pierre (one of the authors of kmindex)