Table 1. Optimal hyperparameters selected by DEWÄKSS self-supervised objective function.
| Normalization | Dataset | iteration | mode | neighbors | pcs | MSE |
|---|---|---|---|---|---|---|
| DESeq2 | celseq2 | 1 | distances | 120 | 3 | 0.466 |
| FTT | celseq2 | 1 | distances | 90 | 5 | 0.878 |
| Linnorm | celseq2 | 1 | distances | 110 | 4 | 0.066 |
| logCPM | celseq2 | 1 | distances | 100 | 6 | 4.567 |
| none | celseq2 | 1 | connectivities | 14 | 120 | 4.428 |
| scone | celseq2 | 1 | distances | 120 | 4 | 0.445 |
| scran | celseq2 | 1 | distances | 130 | 3 | 0.484 |
| TMM | celseq2 | 1 | distances | 50 | 6 | 0.378 |
| DESeq2 | sortseq | 1 | distances | 100 | 3 | 0.513 |
| FTT | sortseq | 1 | distances | 80 | 4 | 1.127 |
| Linnorm | sortseq | 1 | distances | 100 | 4 | 0.083 |
| logCPM | sortseq | 1 | distances | 80 | 13 | 4.684 |
| none | sortseq | 1 | distances | 10 | 17 | 5.321 |
| scone | sortseq | 1 | distances | 100 | 4 | 0.484 |
| scran | sortseq | 1 | distances | 120 | 3 | 0.536 |
| TMM | sortseq | 1 | distances | 50 | 6 | 0.412 |