Methods* | DEG_ori | DEG_rel | DEVG_ori | DEVG_rel | DEG_ori & DEVG_ori | DEG_ori & DEVG_rel |
|---|
Mean of accuracy | 0.7803 | 0.5825 | 0.5965 | 0.6262 | 0.7592 |
0.8871
|
Std of accuracy | 0.0309 | 0.0520 | 0.0229 | 0.0217 | 0.0375 | 0.0918 |
- Italic value indicates the best performance in method comparison.
- * DEG_ori means that we selected genes with differential expression as features, and the original/raw expression values as measurements of these conventional features used in the sample-clustering evaluation; Meanwhile, DEG_rel means that we selected genes with differential expression as features, but the proposed absolute relative expression values as measurements of these conventional features. Similarly, DEVG_ori means that we selected novel genes with differential expression variances as features, but the original/raw expression values as measurements of these new features; DEVG_rel means that we selected novel genes with differential expression variances as features, and absolute relative expression values as suitable measurements of these new features. There are six strategies evaluated, and each strategy applied particular feature genes and corresponding measurements for sample-clustering. For each strategy, the sample–clustering has been rerun 1,000 times, and the mean and variance of accuracies are the final performance of such a strategy.
