Table 1 Experimental conditions affecting differential abundance tests (DATs)

Experimental choices

Number of replicates

NR

Number of technical or non-technical replicates for the two groups compared in the test

For simplicity, in many cases, we assume NR to be the same in both groups

Number of data-points

ND

Number of data-points generated in the counting experiment

For example, reads generated in an RNA-seq experiment

Experimental characteristics

Entity count

EC

Number of entities in the counting experiment

For example, number of genes in an RNA-seq experiment

Sample variability

SV

Variability across replicates (see Methods)

For example, biological variability in RNA-seq datasets

Abundance profile

AP

Relative abundance of the entities in the first group

Typically follows a power-law distribution

Perturbation profile

PDA, FC

Perturbations to the abundance profile of the first group to obtain the profile for the second group (see Methods)

Used to generate the differentially abundant entities (PDA = Percentage of entities, FC = fold-change distribution)

Test settings

Biases in data generation

Deviations from multinomial sampling due to biases inherent in the experimental protocol

These are often corrected for in a preprocessing step, for example, composition bias in RNA-seq data [23],[24]

Differential abundance test

DAT

See Table 2