Table 1.

Median accuracy results obtained from the application of Models I, II, and III to the 1000 bootstrap samples generated (T, tumor neoplastic astrocytes; N, normal astrocytes; EN, elastic net; NB, Naïve Bayes: MSE, mean squared error; AUC, area under the precision-recall curve; Miscl, misclassifications; Vars, nr. of variables selected)

Classes	Model	Vars	Miscl		MSE		AUC
			Train	Test	Train	Test	Train	Test
I - T-core vs. T-periphery	EN	83	10	7	0.029	0.047	0.97	0.94
II - T-periphery vs. N-periphery	EN	85	3	4	0.020	0.037	0.99	0.96
III - T-core/periphery vs. N-periphery	EN	76	1	2	0.005	0.012	0.997	0.982
	Twiner	76	0	2	0.003	0.011	1	0.982
	NBEN	76	5	6	0.009	0.034	0.996	0.979
	NBtwiner	76	4	5	0.008	0.028	0.996	0.981