TABLE 1.

Common characteristics of strategies for dimensionality reduction address different aspects of the data.

Table 1
Term	Definition
Compositionally aware	Transforms data to account for non-independence of features in sequence count data
Pseudo-counts or imputation	Requires no/minimal zeroes in the feature table due to numerical issues (such as logarithm transform being undefined on zeroes)
Able to incorporate phylogeny	Method is calculated with awareness of how each sampled microbial community is evolutionarily represented relative to other samples
Operates on beta-diversity dissimilarities	Dimensionality reduction step is performed on pairwise dissimilarities (arbitrary metric) between samples, rather than the feature table itself
Linear	Lower dimensional coordinates are computed via linear transform of features
Repeated measures	Subjects are sampled multiple times. Commonly sampled longitudinally
Feature relationships are interpretable	The method indicates the relevance of input microbial features with regard to its output coordinates
Supervised component	Method takes explanatory sample variables as an additional input