Table 2. Performance statistic of our three different machine learning algorithms.

Performance statistic during our statistical workflow according to each machine learning algorithm using fresh tumor tissue.

	Logistic Regression	Random Forest	XGBoost
Advantages	simple linear, algorithm for basic differentiation	non-linear algorithm, taken into account the complexity of the spectral signal	strong non-linear, algorithm with possibility of more individual parameter tuning
Performance (balanced accuracy / sensitivity / specificity)	79.3 / 0.91 / 0.68	82.4 / 0.89 / 0.75	76.2 / 0.87 / 0.65
AUC value ROC curve / AUC value PR curve	0.83 / 0.61	0.86 / 0.66	0.79 / 0.53