Taxonomy Analysis questions

taxonomy

(Ruby) #1

Hi,

I’m a total newbie here, and I didn’t took any biology courses before. I was using Qiime 2 to do the taxonomy analysis of these two 16s samples. One is from lung, and the other is from kidney. But the metadata files I got looked weird, because somehow the lung and kidney samples got the exact same categories between their kingdom, phylum, class , order, family, genus and species. I thought there should be some difference. Any idea which step during my Qiime 2 steps went wrong? I’m really stuck here. Any suggestions would be helpful. Thank you!


(Justine) #2

Hi @rubyling,

Welcome!

That the taxonomic listing is the same shouldn’t be a huge concern, IMO. The taxonomic names don’t necessarily specify the actual OTUs in your sample. When I do reference-based clustering, the taxonomy metadata for those samples theoretically cover organisms found in the human gut, dog mouth, the soil outside, and the ocean.

When you make a taxonomy barchart, how do the samples look? Do they have similar composition, or the same organisms?

Best,
Justine


(Ruby) #3

Hi Justine,

Thanks so much for replying my question. When I make the taxonomy barchart in phylum between the lung and kidney samples, they have the exact same composition. (I don’t know if it’s called composition or not, like for phylum, there are p__Proteobacteria, p__Gemmatimonadetes, p__Chloroflexi and etc). I thought they should a least have a little bit differences, like maybe one of the p__ appeared in lung sample but not in kidney sample, but no.

So maybe I did sth wrong when I run import the fastq into Qiime2 or? But they were no bugs when I run the data in Qiime2. Very confused now :frowning: .

Thanks!
Ruby


(Ruby) #4

Hi Justine,

Sorry I forgot to ask sth on my last reply to you. Could you please tell me how to specify the actual OTUs in the samples? Thank you very much!!

Best,
Ruby


(Justine) #5

Hi @rubyling,

Hmm. That’s not great, but it happens. Have you looked at deeper levels? So, does the taxonomy remain constant when you go down to the family or genus level? Remember that phylum is like comparing anything with a spinal cord to anything with a shell.

There is also some possibility that your samples are low enough biomass that anything you’re seeing is reagent contamination. Im not sure up to date on the literature in that field, so if you’re looking and find a good review, it would be awesome if you could share it here.

Its worth going back and checking your import file if you did something wrong. Did you use a manifest? However, if your paths are consistent, and you don’t think there was a mix up earlier in the processing, it’s possibly real.

Im not totally sure what you’re asking. You can export the feature table using qiime tools export. You can visualize every OTU/ASV with qiime feature-table heatmap. Hopefully these are two good options for you.

Best,
Justine


(Ruby) #6

Hmm. That’s not great, but it happens. Have you looked at deeper levels? So, does the taxonomy remain constant when you go down to the family or genus level? Remember that phylum is like comparing anything with a spinal cord to anything with a shell.

There is also some possibility that your samples are low enough biomass that anything you’re seeing is reagent contamination. Im not sure up to date on the literature in that field, so if you’re looking and find a good review, it would be awesome if you could share it here.

Again thanks so much for your answers. I looked at deeper levels, and they are all the same that’s why I’m very confused.

Its worth going back and checking your import file if you did something wrong. Did you use a manifest? However, if your paths are consistent, and you don’t think there was a mix up earlier in the processing, it’s possibly real.

I didn’t use manifest. My original datas are paired end sequences, so I used the "Casava 1.8 paired-end demultiplexed fastq " step to import the data and then I followed: Training feature classifiers with q2-feature-classifier: https://docs.qiime2.org/2018.11/tutorials/feature-classifier/
Then I dragged the taxonomy.qza I got into Qiime View and use that metadata to do the taxonomy barchart.
Maybe this is not the way to do it?


(Justine) #7

Hi Ruby,

Okay, I just wanted to check. Sometimes, the taxonomic comparison can be really similar at, say, phylum level, but you can see more when you go deeper.

Would you mind running qiime diversity beta-diversity on your samples with the bray-curtis and jaccard distance and letting me know the distance between the two samples. You’ll need to rarify the table, first.

That’s good. So, it’s less likely that there was an error at the import. The training sounds fine to me. The resason Im asking is because Im concerned there was a mix up somewhere and the same sample got run twice with different names.

I would suggest going back and double checking your lab notebook, as well as the Illumina sample sheet, if you have access to that. These questions are to make sure you didn’t run the same sample twice, somehow. And, if you did, to figure out where so you can limit what you have to re-do.

My last concern with these samples is that if they’re (Im guessing) biopsy samples from the lung and kidney, they’re probably very low biomass. I would absloutely check your negative controls, to see what (if anything)_shows up there. There’s also discussion of a positive single culture control, so if you have something like that, worthing checking what shows up there. Reagent contamination is a fun and interesting problem in this field, and there have been a lot of discussions on this forum. There’s a big thread here, which is where I think I might start. I dont do a ton of work in low biomass samples, but Im always interested, so if you do a lit review and/or find a good paper, please share it!

Best,
Justine


(Ruby) #8

Hi Justine thank you for your help and advices!

Would you mind running qiime diversity beta-diversity on your samples with the bray-curtis and jaccard distance and letting me know the distance between the two samples. You’ll need to rarify the table, first.

You mean running this step on the dataset I create after the “Casava 1.8 paired-end demultiplexed fastq” ? I will try that first!

I would suggest going back and double checking your lab notebook, as well as the Illumina sample sheet, if you have access to that. These questions are to make sure you didn’t run the same sample twice, somehow. And, if you did, to figure out where so you can limit what you have to re-do.
My last concern with these samples is that if they’re (Im guessing) biopsy samples from the lung and kidney, they’re probably very low biomass. I would absloutely check your negative controls, to see what (if anything)_shows up there. There’s also discussion of a positive single culture control, so if you have something like that, worthing checking what shows up there.

Actually I don’t know much regarding the dataset, I’m a totally newbie in this field. But I will check with my working partner about this and get back to you!


(Justine) #9

Hi @rubyling,

Nope, please take your current table, run rarefaction as deep as you can using qiime feature-table rarefy, and then qiime diversity beta-diveristy.

If you have the negative control samples or blank control samples, it would also be helpful to look at the taxonomy compared to your two biological samples, and the distance between those samples and your biological ones.

Best,
Justine