bugs_list.tsv question

36 views
Skip to first unread message

Paul Naphtali

unread,
Oct 9, 2019, 12:04:25 AM10/9/19
to HUMAnN Users
Hello,
I am currently working with a CF lung metatranscriptome data set that contains 10000-50000 reads per fastq file after removing human and rRNA reads. I ran these fastq files through Humann2 with the following command: 

humann2 --metaphlan ~/.local/bin/metaphlan2 --input ${mysample} --output ./output_directory

When I ran my fastq files through this command, I noticed that metaphlan2 isn't picking up any of the taxa I know to be present when I run these samples through other taxonomic identification programs like KRAKEN and KAIJU. Instead, I receive the following warning: 

Total species selected from prescreen: 0

Selected species explain 0.00% of predicted community composition

No species were selected from the prescreen.
Because of this the custom ChocoPhlAn database is empty.
This will result in zero species-specific gene families and pathways

Since I already had an idea of which species were present in CF sputa using KRAKEN and KAIJU (eg. Pseudomonas aeruginosa, Staphylococcus aureus, Stenotrophomonas maltophilia, Streptococcus spp., etc), I wanted Humann2 to create a custom Chocophlan2 database with those taxa before performing the nucleotide-level search. I added the --taxonomic-profile bugs_list.tsv flag so that a custom Chocophlan database could be prepared from the bugs list I made. 

My issue is that I'm not sure how to format the bugs_list.tsv file since running Metaphlan2 on my samples yielded an output file of 100% unclassified for half of my samples. 

Eric Franzosa

unread,
Oct 9, 2019, 1:16:51 PM10/9/19
to humann...@googlegroups.com
Hi Paul,

Those sequencing depths are pretty low relative to where our methods usually operate. Assuming 100 nt reads, 50K reads = 5 Mnt, which is ~1x coverage of a single bacterial genome. HUMAnN2's sensitivity falls off below 1x coverage, so I'm not sure that you'd get useful results even if you reformatted your existing taxonomic profile for HUMAnN2.

Thanks,
Eric



--
You received this message because you are subscribed to the Google Groups "HUMAnN Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to humann-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/humann-users/18ff3e78-7369-4c10-99d9-8d5b241c02e7%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages