Table 1.

Microbiome sample source tracking methods

Method type	Algorithm	Data pre-processing	Computational model	Computational resource	Representative tool	Ref.
Distance-based	Pair-wise sample distance or similarity	No feature selection	Pair-wise distance calculation	Multi-thread, GPU acceleration	JSD	[52]
					UniFrac	[53]
					Meta-Storms	[54]
					Meta-Prism	[55]
Unsupervised	Bayesian; EM	No feature selection	Model-free unsupervised learning	Multi-thread	SourceTracker	[5]
					FEAST	[6]
Supervised	Ensemble learning; deep learning	Feature selection before source tracking	Model-based supervised learning	Multi-thread, GPU acceleration	Random forest	[56]
					ONN4MST	[61]
					EXPERT	[62]

Note: GPU, graphics processing unit; EM, expectation maximization.