Single-cell classification using graph convolutional networks

BMC Bioinformatics

Table 3 Accuracy of the eight scRNAseq data classifier and the four conventional classifiers tools on the four experiments (N = 1000)

Training dataset	BaronHuman + Muraro + Segerstolpe	Xin + Muraro + Segerstolpe	Xin + BaronHuman + Segerstolpe	Xin + BaronHuman + Muraro
Testing dataset	Xin	BaronHuman	Muraro	Segerstolpe
sigGCN	0.997	0.987	0.974	0.993
FC	0.992	0.977	0.968	0.993
scID	0.989	0.747	0.97	0.979
scPred	0.945	0.467	0.92	0.814
CasTLe	0.99	0.944	0.977	0.992
SingleR	0.995	0.984	0.977	0.996
scmapcluster	0.196	0.003	0.051	0.568
scmapcell	0.756	0.421	0.64	0.367
ACTINN	0.993	0.984	0.974	0.992
RF	0.982	0.941	0.947	0.938
SVM-linear	0.994	0.979	0.972	0.992
SVM-rbf	0.986	0.983	0.973	0.97
KNN	0.934	0.864	0.788	0.817

ISSN: 1471-2105