Will there be a release of HUMAnN that can work with long read data as metaphlan4.2.2 can?
Yes, this is something we’re working on now, but it is a ways off. Probably a HUMAnN v4.5 or (more realistically) v5 feature.
Thanks for letting me know! Using Metaphlan4.2.2 for my long read data has been great.
Hi there, would it be possible to run ONT long read through HUMAnN if you broke down the reads into shorter fragments using a similar approach to the split reads option in metaphlan?
Yes, this ought to work, although I always feel bad suggesting it since it’s then throwing away the advantage of working with long reads!
Thanks for the response! Unfortunately the tools for functional capacity annotation of long read data are few and far between. I can still use my long reads for genome assembly and metaphlan. In regards to the chop my reads ONT reads do you have any suggestions of the QC filtering parameters I should use? I was thinking of bumping up my Q threshold to Q30 and decreasing my read length threshold to 100 bp. I’d love any suggestions you may have! Thanks!
After reading about the compatibility between Metaphlan4.2.2 and Humann3.9 or 4.0alpha is it possible to run Metaphlan4.2.2 with the older database and then feed that taxonomic profile into Humann? I am working with long read data so I cannot downgrade the Metaphlan I am using. My plan is to chop up my reads so they are humann compatible but use the taxonomic profiles generated by Metaphlan4.2.2 using the long read mode. Is this approach feasible? Thanks!
@franzosa I realized that I cannot downgrade the database because the minimap index is essential for the long reads. Is there any way to “trick“ humann3.9 into accepting the Metaphlan4.2.2 +Jan25 output? I have checked which classifications present in my Metaphlan4.2.2 +Jan25 taxonomic profile are missing from the Jun23_202403 or Oct22_202403 databases and made a new taxonomic profile only containing those organisms found in Jun23_202403 or Oct22_202403 but it is unclear to me what my next steps should be. I have also thought about just bypassing the prescreen and getting a community level functional profile but the compute time becomes quite large. Any insight would be helpful. Thanks again!