Table 2:

Average classification accuracy for the rLDA classifier using as input MLP, ECC, or ECP. Data for each tumor are split into 80/20 train–test splits and classification accuracy is reported as the mean over 100 repetitions of splitting, training, and testing.

	CD68+ vs. FoxP3+	CD8+ vs. FoxP3+	CD8+ vs. CD68+	CD8+ vs. CD68+ vs. FoxP3+
	MPL - ECC - ECP	MPL - ECC - ECP	MPL - ECC - ECP	MPL - ECC - ECP
T_A	0.491 - 0.967 - 0.964	0.642 - 0.973 - 0.967	0.630 - 0.840 - 0.830	0.427 - 0.858 - 0.859
T_B	0.760 - 0.892 - 0.869	0.787 - 0.986 - 0.985	0.671 - 0.942 - 0.945	0.604 - 0.868 - 0.865
T_C	0.863 - 0.906 - 0.896	0.747 - 0.847 - 0.842	0.653 - 0.584 - 0.614	0.640 - 0.628 - 0.627
T_D	0.683 - 0.926 - 0.918	0.829 - 0.990 - 0.988	0.476 - 0.779 - 0.779	0.492 - 0.779 - 0.775
T_E	0.820 - 0.886 - 0.883	0.736 - 0.929 - 0.920	0.534 - 0.735 - 0.743	0.502 - 0.702 - 0.683
T_F	0.623 - 0.899 - 0.925	0.476 - 0.842 - 0.847	0.765 - 0.909 - 0.921	0.408 - 0.845 - 0.847
T_G	0.886 - 0.932 - 0.927	0.897 - 0.970 - 0.975	0.446 - 0.696 - 0.692	0.581 - 0.738 - 0.746
T_H	0.524 - 0.890 - 0.898	0.735 - 0.930 - 0.929	0.714 - 0.882 - 0.877	0.502 - 0.844 - 0.859
T_I	0.859 - 0.912 - 0.931	0.883 - 0.908 - 0.909	0.484 - 0.470 - 0.474	0.597 - 0.619 - 0.614
T_J	0.608 - 0.763 - 0.750	0.750 - 0.835 - 0.872	0.850 - 0.882 - 0.892	0.536 - 0.653 - 0.670
T_K	0.376 - 0.868 - 0.804	0.523 - 0.918 - 0.914	0.455 - 0.857 - 0.845	0.261 - 0.718 - 0.679
T_N	0.410 - 0.527 - 0.563	0.432 - 0.662 - 0.745	0.643 - 0.690 - 0.713	0.294 - 0.388 - 0.460
T_O	0.702 - 0.954 - 0.952	0.644 - 0.806 - 0.772	0.546 - 0.672 - 0.684	0.429 - 0.639 - 0.632