Skip to main content
. 2023 Mar 17;39(4):btad133. doi: 10.1093/bioinformatics/btad133

Table 1.

Performance of single clustering of compared methods on different datasets.a

JSNMF UnionCom scMVAE-PoE scMVAE-NN scMVAE-Direct DCCA scMCs
D1 NMI 0.262 ± 0.003• 0.704 ± 0.004• 0.852 ± 0.002• 0.817 ± 0.001• 0.811 ± 0.000• 0.619 ± 0.000• 0.907  ±  0.000
ARI 0.196 ± 0.003• 0.670 ± 0.005• 0.839 ± 0.001• 0.819 ± 0.000• 0.811 ± 0.001• 0.513 ± 0.001• 0.939  ±  0.000
D2 NMI 0.416 ± 0.000• 0.606 ± 0.000° 0.603 ± 0.002° 0.611  ±  0.001° 0.505 ± 0.002• 0.414 ± 0.000• 0.534 ± 0.001
ARI 0.284 ±0.004• 0.400 ± 0.001• 0.452 ± 0.007• 0.447 ± 0.003• 0.441 ± 0.004• 0.404 ± 0.000• 0.596  ±  0.000
D3 NMI 0.140 ± 0.000• 0.334 ± 0.000• 0.331 ± 0.000• 0.294±0.001• 0.265 ± 0.000• 0.433  ±  0.000
ARI 0.087 ± 0.000• 0.250 ± 0.000• 0.260 ± 0.000° 0.232 ± 0.002• 0.250 ± 0.001• 0.260  ±  0.000
D4 NMI 0.269 ± 0.000• 0.305 ± 0.001• 0.325 ± 0.001• 0.287 ± 0.005• 0.273±0.003• 0.296 ± 0.003• 0.510  ±  0.000
ARI 0.194 ± 0.001• 0.248 ± 0.005• 0.268 ± 0.001• 0.164 ± 0.002• 0.125 ± 0.002• 0.197 ± 0.001• 0.554  ±  0.001
a

• /° indicates whether scMCs is superior/inferior to the other method, with statistical significance checked by pairwise t-test at 95% level. The best results are highlighted in bold font.