mmvec with shotgun metagenome+metabolome

[khemlalnirmalkar]

Hi , I am trying to understand the mmvec tool, but a quick question can we use this tool with shotgun based metagenomic data with metabolomic data? If yes, do i need convert them into biom file or is there any other way? Thanks, Khem

[mortonjt]

Yes, this can work with shotgun data, one of the datasets that we used in our publication was a shotgun metagenomics dataset. Yes, everything has to be inputted as a biom format. See https://biom-format.org

On Thu, Jun 18, 2020, 5:23 PM khemlal notifications@github.com wrote:

Hi , I am trying to understand the mmvec tool, but a quick question can we use this tool with shotgun based metagenomic data with metabolomics data? If yes, do i need convert them into biom file or is there any other way? Thanks, Khem

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXPGRYTTTIFJJHNMG4LRXKOVZANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

Hi Jamie @mortonjt , Thank you for an immediate response, Surely then i will give a try with my dataset,

Thanks, Khem

[luzhang321]

Hi :)

So happy to find that someone has the same question as me.

Here is my situation: I convert my species table(metagenomic data from metaphlan after filtering only the interested species) into biom format. (row:taxonomy;column:sample;content:abundance) And I successfully produce a biplot and heatmap via qiime2.

However, I am so confused... my metagenomic input data doesn't provide the sequence information. But according to the paper/readme principle, the model training and testing needs "sequence information". I try to understand, but failed.

Did I do this analysis in the wrong way? Could you solve my confusion?

Many thanks, Lu

[mortonjt]

HI @luzhang321 , I'm a little confused. If you are able to produce biplots + heatmaps, then it sounds like you are able to get the pipeline working, right?

Hmm, that is an interesting edge case. Most qiime2 pipelines assume that the OTU ids and the taxonomies are decoupled. In the case of metaphylan, these two quantities are merged together, but it shouldn't be a problem as long as the ids are unique.

[luzhang321]

Hi @mortonjt

Thanks for an immediate response:)

I think I had a misunderstanding before. I thought the input id must be "ATCG..." type. I thought the conditional probabilities considering the sequence context, A,T,C,G has different effects before. That seems wrong.

Now I understand that as long as the ids are unique, then it works. It doesn't matter which type of data it is, sequence, OTU ID, taxonomy item, even some random unique id. All of them work, Right? I am a beginner, please forgive me if I ask some stupid questions.

Many thanks!

[mortonjt]

Correct.

On Fri, Jun 19, 2020 at 12:05 PM Lulu notifications@github.com wrote:

Hi @mortonjt https://github.com/mortonjt

Thanks for an immediate response:)

I think I had a misunderstanding before. I thought the input id must be "ATCG..." type. Now I understand that as long as the ids are unique, then it works. It doesn't matter which type of data it is, either it is an OTU ID or a taxonomy item. Right?

Many thanks!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-646796962, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXNKC6676JZKFBRPFCDRXOSFBANCNFSM4OCHLGBQ .

[luzhang321]

Hi :)

It's me I found that the same data will produce different results. I run mmvec 2 times. Figure 1 from 1st run Figure 2 from 2nd run

I feel confused again... which one I should trust.

[mortonjt]

The are a couple of things to note.

Runs are not guaranteed to give the exact same results due to the random seed. If you see widely different results, it's likely the algorithm hasn't been run long enough and hasn't reach convergence (see Tensorboard tutorial in the Readme). If you still get different results, I'd pick the one with the smallest loss / cv_rmse.

I'd also recommend running the CF tutorial provided in the tutorial, since that example has been stress tested and should give consistent results.

On Fri, Jun 19, 2020, 3:05 PM Lulu notifications@github.com wrote:

Hi :)

It's me I found that the same data will produce different results. I run mmvec 2 times. Figure 1 from 1st run Figure 2 from 2nd run

I feel confused again... which one I should trust.

[image: image] https://user-images.githubusercontent.com/46122020/85179375-66b13100-b2b3-11ea-88ea-fc4037f46607.png [image: image] https://user-images.githubusercontent.com/46122020/85179403-73ce2000-b2b3-11ea-81f9-7ac369be9378.png

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-646860236, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXPWLTWDYYIDL5IH4WDRXPHKXANCNFSM4OCHLGBQ .

[luzhang321]

The are a couple of things to note. Runs are not guaranteed to give the exact same results due to the random seed. If you see widely different results, it's likely the algorithm hasn't been run long enough and hasn't reach convergence (see Tensorboard tutorial in the Readme). If you still get different results, I'd pick the one with the smallest loss / cv_rmse. I'd also recommend running the CF tutorial provided in the tutorial, since that example has been stress tested and should give consistent results. … On Fri, Jun 19, 2020, 3:05 PM Lulu @.***> wrote: Hi :) It's me I found that the same data will produce different results. I run mmvec 2 times. Figure 1 from 1st run Figure 2 from 2nd run I feel confused again... which one I should trust. [image: image] https://user-images.githubusercontent.com/46122020/85179375-66b13100-b2b3-11ea-88ea-fc4037f46607.png [image: image] https://user-images.githubusercontent.com/46122020/85179403-73ce2000-b2b3-11ea-81f9-7ac369be9378.png — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#135 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXPWLTWDYYIDL5IH4WDRXPHKXANCNFSM4OCHLGBQ .

Hi, Thanks for your quick reply. : ) The command I used is exactly the same as the CF tutorial. I only add the parameters to show the x,y label in the heatmap. I used mmvec through qiime2. The result is widely different.

My microbiome table is only a 12 species table with 30+samples; and the metabolites table has 7 metabolites with 30 samples. Is the data size okay?

The tensorboard tutorail seems complicated. Except manually adjust it, is there a way that I could simply adjust in the qiime2?

Many thanks!

[luzhang321]

Hi, :)

I checked the results again and did 5 loops to compare results. The log probability files from ranks.qza are very similar.

[mortonjt]

Great!

On Sun, Jun 21, 2020, 7:18 AM Lulu notifications@github.com wrote:

Hi, :)

I checked the results again and did 5 loops to compare results. The log probability files from ranks.qza are very similar.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-647127279, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXOV2V3YGBCDWCP4CDTRXYCA7ANCNFSM4OCHLGBQ .

[luzhang321]

Hi @mortonjt

Just follow what you suggested, I read this Tensorboard tutorial in the Readme

And I got the following results.

It's a very weird result, I only have a dot in the plot...

Does that mean my input is wrong?

Thanks!

[mortonjt]

That means you only logged the information once. It maybe because you have a very short run - so you may need to update your summary-interval

On Tue, Jun 23, 2020, 12:01 PM Lulu notifications@github.com wrote:

Hi @mortonjt https://github.com/mortonjt

Just follow what you suggested, I read this Tensorboard tutorial in the Readme

And I got the following results.

It's a very weird result, I only have a dot in the plot...

Does that mean my input is wrong?

[image: image] https://user-images.githubusercontent.com/46122020/85438355-322bc680-b58c-11ea-9621-ae3afaa189d0.png

Thanks!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-648324851, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXNJ6YBXP3VWNCYGPJTRYDUXDANCNFSM4OCHLGBQ .

[luzhang321]

Hi @mortonjt

I am grateful for your quick reply.

Yes. I indeed have a short run. It finished within one minute.

Sorry, I didn't understand it clearly.. what does the "summary-interval" mean?

Thanks very much!

[luzhang321]

Hi :)

I tested it again on the tutorial datasets cf with the command. It is still a dot... So maybe it is because of my environment?

mmvec paired-omics --microbe-file mmvec/examples/cf/otus_nt.biom --metabolite-file mmvec/examples/cf/lcms_nt.biom --summary-dir summary

Thanks.

[mortonjt]

No, it's because you haven't set --summary-interval. Try running mmvec paired-omics --help to see all of the options. Right now it is set to record every 1000 seconds, which is probably too long for your run. Maybe try setting --summary-interval 1 to record every second.

On Wed, Jun 24, 2020 at 6:15 AM Lulu notifications@github.com wrote:

Hi :)

I tested it again on the tutorial datasets cf with the command. It is still a dot... So maybe it is because of my environment?

mmvec paired-omics \ --microbe-file mmvec/examples/cf/otus_nt.biom \ --metabolite-file mmvec/examples/cf/lcms_nt.biom \ --summary-dir summary

[image: image] https://user-images.githubusercontent.com/46122020/85537132-fd635200-b613-11ea-89f9-a7f431474750.png

Thanks.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-648731725, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXNURWLGQXET337SDODRYHG57ANCNFSM4OCHLGBQ .

[luzhang321]

Hi @mortonjt thanks! After setting summary-interval 1, now it is a curve.

I feel confused about the difference of y-axis However, the curve y-axis is not showing like the tutorial(cf). I run cf data as a test.

qiime mmvec paired-omics --i-microbes otus_nt.qza --i-metabolites lcms_nt.qza --p-learning-rate 1e-3 --o-conditionals ranks4.qza --o-conditional-biplot biplot4.qza --p-summary-interval 1

qiime mmvec heatmap --i-ranks ranks4.qza --m-microbe-metadata-file microbe-metadata.txt --m-microbe-metadata-column Taxon --p-level 3 --o-visualization ranks-heatmap4.qzv

Here is the code I used.

Tutorial tensorboard figure :

as the instructions in readme. The x-axis is training time, it is easy to understand. But the y-axis above figure is intensity?

But in my figure followed by tutorial

The y-axis is something else,not the intensity...

Or they use different dataset, I could only focus on if it's plateau, then it means the model is robust? And I can try with my own data?

please forgive me if I asked some stupid questions again.

[luzhang321]

My another question is that I use

biom convert to convert example/cf biom file to tsv table.

I checked the lcms_nt.biom converted tsv and otus_nt.biom converted tsv. And found that they have different columns. But the samples number should be the same, right?

OTUS_nt.biom converted to tsv, has 7559 rows(otus) and 636 samples lcms_nt.biom converted to tsv, has 463 rows(metabolites) and 180 samples

I am confused here. Thanks very much!

[mortonjt]

Right, there are more 16S samples than ms samples. MMvec performs matching internally, so samples that only have one of those datatypes are discarded

On Fri, Jul 3, 2020, 4:19 AM Lulu notifications@github.com wrote:

My another question is that I use

biom convert to convert example/cf biom file to tsv table.

I checked the lcms_nt.biom converted tsv and otus_nt.biom converted tsv. And found that they have different columns. But the samples number should be the same, right?

OTUS_nt.biom converted to tsv, has 7559 rows(otus) and 636 samples lcms_nt.biom converted to tsv, has 463 rows(metabolites) and 180 samples

I am confused here. Thanks very much!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-653471984, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXNUQYVBF2KPKWB43ALRZWWEPANCNFSM4OCHLGBQ .

[mortonjt]

@luzhang321 @khemlalnirmalkar it looks like the original questions have been addressed here, so I'll close this issue. But feel free to follow up if there are outstanding concerns.

[luzhang321]

Hi I also attached this question before... Could you please give me some suggestions if you have time?

Hi @mortonjt thanks! After setting summary-interval 1, now it is a curve.

I feel confused about the difference of y-axis However, the curve y-axis is not showing like the tutorial(cf). I run cf data as a test.

qiime mmvec paired-omics --i-microbes otus_nt.qza --i-metabolites lcms_nt.qza --p-learning-rate 1e-3 --o-conditionals ranks4.qza --o-conditional-biplot biplot4.qza --p-summary-interval 1

qiime mmvec heatmap --i-ranks ranks4.qza --m-microbe-metadata-file microbe-metadata.txt --m-microbe-metadata-column Taxon --p-level 3 --o-visualization ranks-heatmap4.qzv

Here is the code I used.

Tutorial tensorboard figure :

as the instructions in readme. The x-axis is training time, it is easy to understand. But the y-axis above figure is intensity?

But in my figure followed by tutorial

The y-axis is something else,not the intensity...

Or they use different dataset, I could only focus on if it's plateau, then it means the model is robust? And I can try with my own data?

please forgive me if I asked some stupid questions again.

[mortonjt]

Not quite, it's the training loss. This paper may give some insight : https://arxiv.org/abs/1609.04747

On Mon, Jul 6, 2020, 1:46 AM Lulu notifications@github.com wrote:

Hi I also attached this question before... Could you please give me some suggestions if you have time?

Hi @mortonjt https://github.com/mortonjt thanks! After setting summary-interval 1, now it is a curve.

I feel confused about the difference of y-axis However, the curve y-axis is not showing like the tutorial(cf). I run cf data as a test.

qiime mmvec paired-omics --i-microbes otus_nt.qza --i-metabolites lcms_nt.qza --p-learning-rate 1e-3 --o-conditionals ranks4.qza --o-conditional-biplot biplot4.qza --p-summary-interval 1

qiime mmvec heatmap --i-ranks ranks4.qza --m-microbe-metadata-file microbe-metadata.txt --m-microbe-metadata-column Taxon --p-level 3 --o-visualization ranks-heatmap4.qzv

Here is the code I used.

Tutorial tensorboard figure :

[image: image] https://user-images.githubusercontent.com/46122020/86458501-9df8f680-bd25-11ea-952f-4ef782e7648f.png

as the instructions in readme. The x-axis is training time, it is easy to understand. But the y-axis above figure is intensity?

But in my figure followed by tutorial

[image: image] https://user-images.githubusercontent.com/46122020/86458916-50c95480-bd26-11ea-9068-0c97319bab9a.png

The y-axis is something else,not the intensity...

Or they use different dataset, I could only focus on if it's plateau, then it means the model is robust? And I can try with my own data?

please forgive me if I asked some stupid questions again.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-654073370, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXMQWQI7S4SNSOU7WADR2F6NTANCNFSM4OCHLGBQ .

[luzhang321]

Hi thanks for your quick reply and suggestions for the paper.:)

But the data I used is the tutorial cf data. In theory, it shouldn't look like that.. I didn't use my own data.

Thanks very much!

[mortonjt]

Got it, I think I see where the confusion is.

The tensorboard figure in the README was applied to metabolites after being normalized -- so the metabolites are proportions and the cv_rmse represents the differences in proportions.

If you run the CF tutorial, that won't normalize the metabolites and that'll give you the differences in intensities. This is expected to be large (since intensities are typically 10^8).

Since log-loss is also impacted by normalization / model parameters we don't expect that to be the same, and the most important detail to focus on is whether or not the model has reached convergence.

Best, Jamie

On Mon, Jul 6, 2020 at 11:52 AM Lulu notifications@github.com wrote:

Hi thanks for your quick reply and suggestions for the paper.:)

But the data I used is the tutorial cf data. In theory, it shouldn't look like that.. I didn't use my own data.

Thanks very much!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-654379628, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXO5F7RPRRLM2JI5OULR2IFMPANCNFSM4OCHLGBQ .

[luzhang321]

aha! Got it, thanks so much!

[khemlalnirmalkar]

Hi :)

It's me I found that the same data will produce different results. I run mmvec 2 times. Figure 1 from 1st run Figure 2 from 2nd run

I feel confused again... which one I should trust.

Hi @luzhang321 ....did you make your heatmap using mmvec's python script (e.g. like Figure6-inflammatory-bowel-disease.ipynb)? creating ordination and modifying the feature data seems very complicated. I am not good at python. Is there any simple way? please can you share your experience?

Hi @mortonjt since mmvec has qiime2 plugin, is there any simple way to make a heatmap for microbe-metabolites interaction like how we run other steps in qiime2, without using python directly? Sorry, "Figure6-inflammatory-bowel-disease.ipynb" has a lot of steps which i am not able to follow. A newbie like me for python, simple way would be very helpful, Thanks in advance,

[mortonjt]

yes - see the qiime2 plugin section, we have a qiime mmvec heatmap command.

[khemlalnirmalkar]

Hi @mortonjt , Thanks, i checked the section, last time when i checked, I missed the row of compound source in the heatmap (not visible in the github image), and that's why i got the doubt but when i ran in qiime, i could see now. One more doubt, is that okay to use the relative abundance of the microbiome (shotgun) and metabolomics data (imputed and normalized) in z-score? z-score based metabolomics (untargeted) data have negative values, will it affect the run and results? as i understood, your sample files in the tutorials are the absolute abundance for microbes and metabolites data are Unnormalized. I will look forward to your suggestion. Thansks, Khem

[mortonjt]

No, z scores will not work. Normalized values are OK, but no guarantees. Best to feed in raw abundances if you can.

On Thu, Sep 3, 2020, 3:39 PM khemlal notifications@github.com wrote:

Hi @mortonjt https://github.com/mortonjt , Thanks, i checked the section, last time when i checked, I missed the row of compound source in the heatmap (not visible in the github image), and that's why i got the doubt but when i ran in qiime, i could see now. One more doubt, is that okay to use the relative abundance of the microbiome (shotgun) and metabolomics data (imputed and normalized) in z-score? z-score based metabolomics (untargeted) data have negative values, will it affect the run and results? as i understood, your sample files in the tutorials are the absolute abundance for microbes and metabolites data are Unnormalized. I will look forward to your suggestion. Thansks, Khem

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-686778049, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXJ6RDZGXXYGCM6Z25LSEAEJPANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

Okay, thanks for the quick response,

[khemlalnirmalkar]

Hi @mortonjt , I made this heatmap for microbe-metabolites interaction but since i have 3 study groups (case, control, treatment, see another attached image ), i am interested to add all 3 groups in a single heatmap to see the microbe-metabolites interaction. But "qiime mmvec paired-omics" gives rank.qza file only for microbe-metabolites without groups. Please may i know how can i add all 3 groups in a single image without losing the microbe-metabolites interaction? is it possible in mmvec?

Thanks, Khem

[mortonjt]

You probably need to manually specify your training / testing examples. MMvec does this automatically and there is a chance that one of your groups got filtered out because of this.

On Fri, Sep 4, 2020, 3:22 AM khemlal notifications@github.com wrote:

Hi @mortonjt https://github.com/mortonjt , I made this heatmap for microbe-metabolites interaction but since i have 3 study groups (case, control, treatment, see another attached image [image: example] https://user-images.githubusercontent.com/32944151/92223239-06af3d00-ee55-11ea-92ab-2126b2b7267c.JPG ), i am interested to add all 3 groups in a single heatmap to see the microbe-metabolites interaction. But "qiime mmvec paired-omics" gives rank.qza file only for microbe-metabolites without groups. Please may i know how can i add all 3 groups in a single image without losing the microbe-metabolites interaction? is it possible in mmvec?

Thanks, Khem

[image: heatmap] https://user-images.githubusercontent.com/32944151/92221518-7839bc00-ee52-11ea-9d06-e6bb5ebd9c37.JPG

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-687032147, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXPKBVOIBCDJJW4MHF3SECWTZANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

Thanks @mortonjt , I was checking the options for manual adjustment for 3group through mmvec, i couldn't find the options in qiime mmvec heatmap --help. p-method and p-metric parameters are for the clustering method, but i may need to stop clustering on top /clustering_column in order to have 3 groups. Any tip? Khem

[mortonjt]

sorry, I misunderstood your previous question.

You're asking about how to plot samples in your heatmap right? The qiime mmvec heatmap command only plots the interactions -- but the qiime mmvec paired-heatmap command can help with this.

If you are still having trouble, including your command will help me provide better feedback.

On Fri, Sep 4, 2020 at 1:44 PM khemlal notifications@github.com wrote:

Thanks @mortonjt https://github.com/mortonjt , I was checking the options for manual adjustment for 3group through mmvec, i couldn't find the options in qiime mmvec heatmap --help. p-method and p-metric parameters are for the clustering method, but i may need to stop clustering on top /clustering_column in order to have 3 groups. Any tip? Khem

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-687345814, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXIOCKT4RGU2Y7BXEWDSEE7QNANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

@mortonjt , I guess previously you understood correctly, my apologies if I am not explaining myself properly. As I mentioned before I have 3 groups (each group 16 samples), what I wanted is, to have all 3 groups in the same heatmap showing the interaction between microbe-metabolites. So readers can see the difference between the group. Something similar to this attached image, Khem

[mortonjt]

I'm not seeing how interactions are represented in your heatmap...

Does the paired heatmap address your concerns?

On Fri, Sep 4, 2020, 2:42 PM khemlal notifications@github.com wrote:

@mortonjt https://github.com/mortonjt , I guess previously you understood correctly, my apologies if I am not explaining myself properly. As I mentioned before I have 3 groups (each group 16 samples), what I wanted is, to have all 3 groups in the same heatmap showing the interaction between microbe-metabolites. So readers can see the difference between the group. Something similar to this attached image, [image: example] https://user-images.githubusercontent.com/32944151/92283256-5d982f00-eeb4-11ea-8feb-bd92b3413eb5.JPG Khem

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-687372685, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXNRABPCAZJBKQ4IKPLSEFGLRANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

Sorry, this is just an example to show how i want my 3 groups in the column, nothing about interaction. now i am trying paired heatmap option...but i got an error...below is my command,

qiime mmvec paired-heatmap \ --i-ranks ranks.qza \ --i-microbes-table taxon_biom.qza \ --i-metabolites-table metabolites_biom.qza \ --m-microbe-metadata-file taxon_metadata.txt \ --m-microbe-metadata-column Taxon \ --p-features Bac1 \ --p-features Bac2 \ --p-top-k-microbes 0 \ --p-normalize rel_row \ --p-top-k-metabolites 100 \ --p-level 6 \ --o-visualization paired-heatmap-top2.qzv Plugin error from mmvec:

cannot do a non-empty take from an empty axes.

Debug info has been saved to /tmp/qiime2-q2cli-err-4mttpbhm.log

Now i am trying to find whats wrong with my input file

[khemlalnirmalkar]

Hi @mortonjt 2 concerns

1. This paired heatmap doesnt show the name of metabolites. I couldnt find the options in help section
   2. Couldn't find the way to show all 3 groups, Please let me know if you have any more suggestions. Thanks,

[mortonjt]

There is a chance that we can't do this visualization in qiime2 -- it's a bit too specialized.

But you can always read the ranks file as a text file via qiime tools export : https://docs.qiime2.org/2020.8/tutorials/exporting/ So you can directly read off the ranks / log conditional probabilities in whatever interface you like.

On Fri, Sep 4, 2020 at 5:28 PM khemlal notifications@github.com wrote:

Hi @mortonjt https://github.com/mortonjt 2 concerns

1. This paired heatmap doesnt show the name of metabolites. I couldnt find the options in help section
2. Couldn't find the way to show all 3 groups, Please let me know if you have any more suggestions. Thanks,

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-687465538, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXMWE4HYHSRSJIFDHR3SEFZ2DANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

Okay Thanks @mortonjt , I will try

[khemlalnirmalkar]

Hi @mortonjt , One quick question, is it safe to say, positive log conditional probabilities for microbe (x)-metabolite (y) co-occurrence indicate that metabolite (y) may be produced or positively regulated by microbe (x)? i am just trying to find a way simple way to explain the mmvec heatmap and biplot. Thanks,

[mortonjt]

Yes, that is a valid hypothesis, but that would need some follow up, since we cannot claim causality here.

On Mon, Sep 7, 2020, 1:19 PM khemlal notifications@github.com wrote:

Hi @mortonjt https://github.com/mortonjt , One quick question, is it safe to say, positive log conditional probabilities for microbe (x)-metabolite (y) co-occurrence indicate that metabolite (y) may be produced or positively regulated by microbe (x)? i am just trying to find a way simple way to explain the mmvec heatmap and biplot. Thanks,

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/biocore/mmvec/issues/135#issuecomment-688481463, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA75VXLMKFGRXEZAA7ID6KTSEUW4ZANCNFSM4OCHLGBQ .

[khemlalnirmalkar]

Okay, Thanks @mortonjt , I appreciate your kind support, especially your quick response has been very helpful to find the solution immediately, Good day,