I have a question about if it's possible to make minor changes in a metadata file without changing their results. I tried to change the text of a column in metadata file, like this:
==> metadata.tsv <==
id patient status type_sample n_bacteria_sorted
S1 V1.UC_07-1 Control PBS_1 50000
S2 V1.UC_07-1 Control P2_(Input) 50000
S3 V1.UC_07-1 Control P3_(IgA1) 19000
...
==> metadata_clear-sample.csv <==
id patient status type_sample n_bacteria_sorted
S1 V1.UC_07-1 Control PBS 50000
S2 V1.UC_07-1 Control Input 50000
S3 V1.UC_07-1 Control IgA1 19000
...
In bold, the changes I wanted to do (only text).
The problem, is that if I run "qiime diversity core-metrics-phylogenetic" with each metadata file I'm obtaining different plots, for example in jaccard_emperor.
I don't understand those changes, and if this is the expected behaviour.
Metadata is in some ways meant to be mutable. Over the course of an experiment we obtain new data, find mistakes in the old data, and make new metadata categories/values to accommodate the needs of our analysis. So go ahead and change your metadata as needed!
Obviously this can change your results if changing your metadata values results in new groupings of data... but as in your case (where your 3 metadata values are being given 3 new unique labels), some diversity estimate results can still change. This is because some analysis rely on random sampling, as in core-metrics-phylogenetic where your input feature table is being randomly subsampled (rarefied) at even depths per sample. Do this 10 times and you will receive slightly different results... if you are subsampling at a reasonably high depth, those results should converge, but if you are subsampling at too low of a depth then those results can be more divergent.