Table 3.

Performance of ADTD, EPIC, CIBERSORTx, and Scaden on the breast cancer test data.^a

	ADTD	EPIC1	EPIC2	CIBERSORTx	Scaden
B-cells	0.824 ± 0.012	0.035 ± 0.022	0.717 ± 0.013	0.315 ± 0.089	0.657 ± 0.018
Endo.	0.832 ± 0.008	0.631 ± 0.011	0.783 ± 0.010	0.633 ± 0.060	0.859 ± 0.010
Myel.b	0.928 ± 0.003	0.867 ± 0.005	0.748 ± 0.012	0.798 ± 0.044	0.794 ± 0.009
Epith.c	0.349 ± 0.022	0.281 ± 0.023	-	0.201 ± 0.092	0.378 ± 0.030
PVLd	0.907 ± 0.005	0.573 ± 0.014	-	0.670 ± 0.062	0.838 ± 0.007
T-cells	0.882 ± 0.007	0.732 ± 0.018	0.419 ± 0.035	0.427 ± 0.096	0.732 ± 0.015
Mean	0.787 ± 0.006	0.520 ± 0.003	0.666 ± 0.010	0.508 ± 0.035	0.710 ± 0.005

^aObserved Pearson’s correlations obtained by comparing the estimated cellular proportions of the cell types captured in the reference matrix X with the ground truth for artificial cellular mixtures of the breast cancer test dataset, where solely the cancer epithelial cells were hidden in the mixtures and not represented as reference profiles (see also Table 4). The error bars correspond to ±1 SD obtained over 10 simulation runs and correlations > 0.8 are highlighted in bold.

^bMyel., myeloid cells.

^cEpith., epithelial cells.

^dPVL, perivascular-like cells.