search

epi2me-labs / wf-16s

Other
22 stars 5 forks source link

Ways to improve minimap classification? #28

Closed osvatic closed 1 month ago

osvatic commented 2 months ago

Ask away!

I am currently trying to use minimap on a ONT dataset to get a finer resolution, ideally down to species, on some full length amplicons. When I run this dataset using kraken as the classifier (with any database) I get a high percentage of the reads classified to some degree. When I run it with minimap I end up with ~99% unclassified regardless of dataset.

Are there settings that could be tweaked to adjust this? Maybe min_percent_identity?

nggvs commented 2 months ago

Hi @osvatic , Thank you very much for using the workflow! You can update to the latest version1.3.0. There is a new table that shows the unmapped sequences and there is also the number of unclassified in the rest of plots. Hopefully, this may give you an idea on how the min_percent_identity and min_ref_coverage are working and you can decrease this value if you consider it appropriate. To use the species level, you can change the taxonomic resolution to "S" (except in the SILVA database). Hope you find this useful!

nggvs commented 1 month ago

Hi @osvatic , Are you still experiencing this issue? Thank you very much in advance!

osvatic commented 1 month ago

No, I was able to use the options to try and adjust the settings to improve the results.