Table 4.

Representative methods for the analysis of microbial dark matter

Microbial dark matter	Representative traditional method	Representative AI method	Summary
Context-dependent biomes	SourceTracker [5], FEAST [6]	ONN4MST [61], EXPERT [62]	AI methods are especially suitable for source tracking among thousands to millions of samples in a fast and accurate manner
Domains of species	QIIME2 [74], Virfinder [76], OrthoDB [79]	\	Current methods for bacteria, archaea, virus, and protist analyses are limited to identifying known species
Functional genes	HUMAnN2 [40], antiSMASH [42]	DeepARG [43], HMD-ARG [121]	Current methods could identify novel genes, but with low speed and low fidelity
Dynamic ecological and evolutionary patterns	PCoA, MITRE [120]	\	Current methods are not sensitive to identifying the dynamic ecological and evolutionary patterns

Note: AI, artificial intelligence; “\”, not reported.