Table 2 FDRs of scFSNN and five other classifiers on six simulated scRNA-seq datasets with different proportions of DE genes. False Discovery Rate (FDR) represents the proportion of features identified as statistically significant but actually irrelevant to the response, among all discovered features. Here, we report the average FDR across 20 replicate experiments. Standard errors are shown in parentheses

0.02

0.1281

0.0901

0.8962

0.9800

0.5579

0.5137

(0.0753)

(0.0523)

(0.1432)

(0.0015)

(0.4372)

(0.3881)

0.03

0.0537

0.0733

0.7996

0.9698

0.3956

0.3358

(0.0394)

(0.0381)

(0.2211)

(0.0019)

(0.4145)

(0.4077)

0.05

0.0368

0.0727

0.6859

0.9506

0.3738

0.1879

(0.0157)

(0.0279)

(0.2884)

(0.0032)

(0.4192)

(0.3518)

0.1

0.0283

0.0881

0.6944

0.8993

0.3039

0.1190

(0.0160)

(0.0223)

(0.2705)

(0.0039)

(0.4110)

(0.2934)

0.2

0.0267

0.0846

0.5507

0.8074

0.2664

0.0868

(0.0181)

(0.0194)

(0.2008)

(0.0057)

(0.3535)

(0.2413)

0.3

0.0281

0.0727

0.4791

0.7201

0.1698

0.1192

(0.0149)

(0.0112)

(0.2202)

(0.0048)

(0.2760)

(0.2567)