Table 4.

Metrics obtained for the random forest model on different datasets. The metrics were computed both in cross-validation (CV) on the train set $\mathcal{T}$ (mean with 95% confidence intervals) and in out-of-sample evaluation on the test set $\mathcal{V}$. In bold, the best performer.

Dataset	ACC CV (CI)	ACC Test	MCC CV (CI)	MCC Test
AllCpGs	0.713 (0.676–0.747)	0.756	0.435 (0.359–0.502)	0.538
ImmuneAngioICIs	0.7155 (0.679–0.754)	0.716	0.436 (0.368–0.512)	0.523
ImmuneAngioICIsMesECM	0.710 (0.674–0.748)	0.739	0.429 (0.354–0.504)	0.490
AllCpGs + BORUTA	0.736 (0.699–0.770)	0.755	0.478 (0.404–0.547)	0.532
ImmuneAngioICIs + BORUTA	0.717 (0.681–0.752)	0.729	0.443 (0.373–0.511)	0.469
ImmuneAngioICIsMesECM + BORUTA	0.747 (0.713–0.780)	0.793	0.498 (0.432–0.563)	0.589