Structural zeros in ANCOM2


For our two-period, placebo-controlled, crossover dietary intervention study, we used our log-ratio transformed abundance data in ANCOM2 to detect differentially abundant taxa. During data preprocessing, 10 taxa were identified as structural zeros that were present in the placebo group but not in the treatment group. Similarly, 8 taxa were identified as structural zeros in the treatment group but not in the placebo group. According to Kaul et al. (2017) , these 18 taxa are then automatically considered to be differentially abundant across the two groups. For publication purposes, may we therefore imply these structural zeros indicate a treatment effect? We also found taxa that were differentially abundant, but were not structural zeros.

Any assistance would be much appreciated!