Table 1.
Comparison of CpG methylation frequencies from whole-genome bisulfite sequencing (WGBS) and from Cas9-targeted nanopore data.
| N | r | r2 | ρ | RMSE | |
|---|---|---|---|---|---|
| Nanopolish | 1704 | 0.8652 | 0.7485 | 0.8326 | 0.2248 |
| DeepSignal | 1731 | 0.9177 | 0.8423 | 0.8765 | 0.1708 |
| Megalodon | 1723 | 0.9117 | 0.8312 | 0.8801 | 0.1772 |
| Tombo | 1661 | 0.7765 | 0.6030 | 0.7537 | 0.2996 |
| Guppy | 1738 | 0.8513 | 0.7246 | 0.8316 | 0.2334 |
| DeepMod | 1739 | 0.7401 | 0.5477 | 0.7264 | 0.2874 |
| METEORE (RF) | 1723 | 0.9174 | 0.8416 | 0.8862 | 0.1829 |
| METEORE (REG) | 1723 | 0.9262 | 0.8579 | 0.8885 | 0.1607 |
For each method, we give the number of sites (N), the Pearson’s correlation (r), coefficient of determination (r2), the Spearman’s rank correlation (ρ), and the root mean square error (RMSE) for the comparison of the percentage methylation predicted from nanopore with the percentage methylation calculated from WGBS Illumina data. We show the results for the six tested tools and METEORE combining Megalodon and DeepSignal using a random forest (RF) (parameters: max_depth = 3 and n_estimator = 10) or a regression (REG) model.