Table 1 FDRs of scFSNN and five other classifiers on five simulated scRNA-seq datasets with different sample sizes. 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
From: scFSNN: a feature selection method based on neural network for single-cell RNA-seq data
Sample size | scFSNN | SurvNet | \(\textrm{L}_1\) | \(\textrm{L}_2\) | GL | SGL |
|---|---|---|---|---|---|---|
1000 | 0.0437 | 0.0757 | 0.6897 | 0.9505 | 0.4619 | 0.4816 |
| Â | (0.0300) | (0.0205) | (0.2789) | (0.0030) | (0.4238) | (0.4527) |
2000 | 0.0359 | 0.0775 | 0.6357 | 0.9505 | 0.5329 | 0.2795 |
| Â | (0.0201) | (0.0336) | (0.3315) | (0.0032) | (0.4261) | (0.4181) |
3000 | 0.0381 | 0.0620 | 0.6328 | 0.9505 | 0.5817 | 0.3005 |
| Â | (0.0277) | (0.0291) | (0.3207) | (0.0031) | (0.3898) | (0.4175) |
4000 | 0.0451 | 0.0748 | 0.6671 | 0.9505 | 0.5093 | 0.2617 |
| Â | (0.0290) | (0.0317) | (0.2503) | (0.0031) | (0.4179) | (0.3898) |
5000 | 0.0334 | 0.0686 | 0.6611 | 0.9505 | 0.4942 | 0.3759 |
| Â | (0.0299) | (0.0307) | (0.3265) | (0.0031) | (0.4417) | (0.4064) |