Normalize or Trensform?

Hello everyone, I have suome doubts about which is the best approach to treat the counts before run the model, normalization or transformation? my feature table has absolute abundances counts which I would like to treat it as compositional. Any suggestions?

Hi @Alessandro_Atzeni,

We don’t normally suggest parameters as it really does depend on your specific data. The MaAsLin 2 manuscript shows some of the parameter tuning that we did to get to the default parameters included in MaAsLin (TSS normalization and log transform), which does tend to work well on most data. The tutorial was also recently expanded to give some guidance on when you can and can’t use various normalization and transformation methods. I hope this helps!