Hi, I am interested to adjust the alpha diversity using my clinical covariates. Can I adjust alpha diversity by placing covariates as fixed effects in Maaslin2, like how I normally do with ASVs? I am thinking no transformation and no normalization in this case.
Hi @jjsimz - I think it’s possible but we have not tested it explicitly. Do let us know if you come across any issues when you have a single feature (i.e., alpha diversity) in your data?
Hi @himel.mallick - I cannot seem to change the correction method to bonferroni. This may be needed as I have only 1 feature. I am using Maaslin2 ver 1.4.0. Otherwise, the pval, when unadjusted, look rather comparable to a standard Mann-Whitney test.
Hi @jjsimz - in general, you should be able to apply any correction method supported by the p.adjust() function which is what is used under the hood in MaAsLin 2. Did correction = "bonferroni" not work in your case?
Hi @himel.mallick - for some reason, correction = "bonferroni" did not work. I will double-check.
It suddenly strikes me that since I only have 1 feature, would it make sense to use pval instead of qval to report my results? I do notice that the pval changes when covariates are added in the fixed_effects, so I am guessing that the covariate adjustment occurs at the level of pval?
Hi @jjsimz - for a single feature and a single covariate, correction is meaningless but MaAsLin 2 should run without error and at least produce identical p- and q-values. For multiple covariates, the adjusted p-values are likely to differ from their unadjusted versions and you may not need correction either given that you only have a single feature.
Hi @jjsimz - the general rule of thumb is that you don’t need multiplicity correction when dealing with one outcome (in your case, alpha diversity). I would personally tend towards reporting p-value for the same reason but I am not against reporting q-value especially when the conclusions are more or less similar (which seems to be the case in your analysis).