Troubleshooting - can't produce a PLSDA, but all else works

pre-treated_Woodchester_blood_TESTx3.csv (3.5 MB)
This data set which I have previously worked fine, will now no longer will run a PLSDA. I’m uploading it (as it is here) as a CSV file under the ‘statistical analysis (one factor)’ tab, with the samples in columns unpaired. I’m selecting no filtering, but am performing log transformation and pareto scaling - at this point the data looks fine and will run heatmaps, PCA etc. But the PLSDA tab will not run and sends me to the ‘Generic Error Landing Page’. Whats the problem?


A quick scan shows that your data contains two factors (test result and sex )- this is not compatible with this module (one factor).

Thanks Jeff, but it does the same with ‘test result’ removed. And in fact I can still run all the other modules in metaboanalyst (PCA, heatmap etc) with ‘test result’ in there, it simply classes it as another sample. It’s just PLSDA that doesn’t work - it spends a long time loading, and then gives me the generic error landing page - a timing out issue perhaps?

I tested your data (after removing “test result”), and didn’t see the issue. Could be that the server was busy at the time. Additionally, I would suggest to apply filtering as untargeted metabolomics data usually contains a lot of noises; and PLS-DA tends to overfit