Hej,
Love your tool and have it used a lot, also in several publications. I am now preparing for a new article. I had previously run DIABLO on metagenomics (taxa) and metabolomics and the plotIndiv after block.plsda looked as the two blocks to the left. I have 9 blue and 9 orange samples and the numbers seem to add up. However, we now had to exclude some metabolites and I am running the same code again, foru months later but with the same mixomics version, and I get the two blocks to the right, where clearly, samples are missing in Block: taxa. I haven’t changed anything consiously, except that I now have 129 instead of 178 metabolites. How can I fix this? Also, I am not sure whether this is a problem of the plotIndiv or something further up and it just shows in plotIndiv.
I very much appreciate your help!
Cheers,
Stef
Also, if this helps: I previously ran
MyResult.diablo1 ← block.plsda(X, Y, ncomp=8)
perf.diablo = perf(MyResult.diablo1, validation = ‘loo’, nrepeats=20, cpus=6)
But now, I get:
Error in checkForRemoteErrors(val) :
one node produced an error: system is computationally singular: reciprocal condition number = 1.47125e-32
and I have to decrease ncomp to 3 to make it run. So that is a new issue upstream.
hi @stepra ,
I am not sure what is happening, unless you have missing values (some individuals not represented?) for the new subset of metabolites?
This issue has never been reported before.
Are you able to extract the variates from the DIABLO object and identify your 18 individuals on both comps 1-2? I wonder if this is a plotting issue from plotIndiv.
Kim-Anh
I’d recommend you create a new post for this as this discussion is going to get messy. You only have 2 groups, so I’d go up to 3 components anyway. By comp > 3 the matrices become ‘empty’ through the deflation step.
Kim-Anh
And I have no misssing values anywhere in my input.
I also just reran my old code not changing anything and I get the same problem in plotIndiv. But the problem does not seem to be the plotting as my models are different now 
@kimanh.lecao I have now installed the latest Omix devtools version and still the same issue. Missing samples in taxa and the models based on the original data are not reproducible. I also tested the code on a Windows machine (I have Mac) with the latest Bioconductor version and get the same results. I am feeling stressed as this is the last analysis we need to rerun for the manuscript and quite a lot depends on it. What can we do?
Hi Stepra, I’m not sure if this is related to your plotting issue, but this error could be related to parallel execution of perf on 6 cpus, do you get this error when you run the same command in serial mode?
Hej Eva, thank you very much for your input. As mentioned above, this issue disappears when I decrease the ncomp but that does not solve the issue of the plotting, and most importantly, of not being able to replicate the models! But it wasn’t there before either.
Dear Kim and Eva,
Thank you for trying to help figure things out! While I was preparing some data to send to Eva, I discovered that some sample names hade empty spaces at their end while others did not and this caused my downstream problems. I could now reproduce and update my analysis and we can move on with our manuscript 
Again, thank you for a great tool!
Stef