Table 1.

Evaluation of clustering results of four real scRNA-Seq datasets.

Metric	Dataset	Raw	AutoClass	DCA	MAGIC	scImpute	SAVER
ARI	Buettner	0.023	0.372	0.288	0.213	0.039	0.016
	Usoskin	0.221	0.869	0.234	0.813	0.067	0.317
	Lake	0.403	0.557	0.572	0.440	0.313	0.465
	Zeisel	0.737	0.793	0.753	0.433	0.623	0.763
JI	Buettner	0.242	0.409	0.363	0.368	0.262	0.247
	Usoskin	0.324	0.830	0.284	0.764	0.266	0.351
	Lake	0.323	0.439	0.453	0.346	0.254	0.364
	Zeisel	0.646	0.713	0.664	0.370	0.679	0.677
NMI	Buettner	0.035	0.395	0.333	0.335	0.075	0.038
	Usoskin	0.225	0.829	0.253	0.771	0.048	0.431
	Lake	0.611	0.667	0.676	0.601	0.500	0.642
	Zeisel	0.747	0.784	0.746	0.598	0.798	0.762
PS	Buettner	0.434	0.720	0.648	0.599	0.445	0.423
	Usoskin	0.545	0.937	0.579	0.913	0.416	0.682
	Lake	0.723	0.772	0.766	0.693	0.610	0.742
	Zeisel	0.894	0.917	0.880	0.763	0.548	0.897

The four metrics are adjusted Rand index (ARI), Jaccard Index (JI), normalized mutual information (NMI), and purity score (PS). Highest value in each row was highlighted in boldface.