Dear all,
now that I know how to handle Qurro, it is time for testing the statistical significance of differences in log ratios between features. Qarcoal calculates the log ratio, but how to perform the t-test? In an old post it was advised to : “once you have a log-ratio of biological interest you can verify it by exporting the metadata with a new log-ratio column and then use a t-test between your groups of interest.”
This is my question: how to do this?

Qurro doesn't currently support performing statistical tests within the visualization (for a few reasons). If you'd like to do a t-test or another significance / effect size / etc. test, you can use the Export current sample plot data button (highlighted in the screenshot below) to save a TSV file containing the current log-ratio values for each sample in the sample plot.

You can then load this TSV file into R, Python, or even something like Excel -- whatever statistics software you prefer -- to run your preferred statistical test. (It's worth noting that the test you choose may be different depending on what sort of sample metadata you're using in the sample plot -- for example if the x-axis is set to a Categorical scale then you may want a t-test, but if the x-axis is set to a Quantitative scale (as in the screenshot above) then you may want to use another test (e.g. a Pearson/Spearman/etc. correlation, or some sort of regression).

If you would like an example of doing this in practice, here's a sort of complicated Jupyter Notebook where we ran linear regression on two TSV files output from Qurro. We used these regressions to get R^2 values (effect sizes) for these datasets. (I generally prefer calculating effect sizes over p-values, but the decision of what test to run is left up to you.)

Hope this helps! I'm tagging @cmartino (the author of the quote you mentioned) in case he has any additional thoughts on this (and to double-check I'm not messing up the statistics here ).