upgrade ganon recipe to use version 0.2.3

This PR updates the ganon used in the lemmi evaluation to the version 0.2.3. It also includes better control over parameters to build reference indices and classification. The Taskfile is now forcing ganon build to create the index at species level (--rank species). It also forces --offset 2 on classification, to speed up analysis.

The following configurations are suggested:

build_ref

parameter=19,240000 for RefSeq/08.2018/All. 19 stands for k-mer size, 240000 maximum memory in MB. I am not sure ganon will be able to produce an index with this amount of memory for this set. Please skip this reference set if it does not work.

parameter=19,96000 for RefSeq/08.2018/1rep. 19 stands for k-mer size, 96000 maximum memory in MB.

analysis

parameter=0.4,single

parameter=0.4,paired

  • the first parameter will set the --min-kmers on ganon classify, which is the threshold for classification. On the current version running on lemmi, ganon is set to use a hard error rate with --max-error 4 which has similar effect than --min-kmers 0.4, but it produces very different results for the sets with longer reads.
  • the second parameter defines the use of the first read file only (single) or both (paired)

It would be interesting to see results with both, since ganon performed well with only the forward reads. However, if only one is allowed, please go on with the paired reads.

Ganon code changed a lot in this version compared to the old one used in lemmi. The parameters used now are also set slightly different so I expect to see changes in results (hopefully better). Please let me know if something is wrong or can be improved.

Merge request reports

Loading