Table 1.

Composite Scores (CS) of survival predictions on adult and pediatric glioma cohorts using two single modality models (FFPE & RNA), and three data fusion methods (early, late & joint).

Cohort	Strategy	Model	Training set CV ean(CS)  ± stdev	Validation set CV Mean(CS)  ± stdev	Test set CS
Adult	Single modality	FFPE	0.874 ± 0.016	0.807 ± 0.018	0.805
		RNA	0.861 ± 0.022	0.809 ± 0.037	0.780
	Multimodal	Early Fusion	0.876 ± 0.008	0.831 ± 0.022	0.836
		Late Fusion	0.891 ± 0.011	0.832 ± 0.019	0.822
		Joint Fusion	0.876 ± 0.026	0.824 ± 0.019	0.822
Pediatric	Single modality	FFPE	0.900± 0.010	0.792 ± 0.070	0.854
		RNA	0.900 ± 0.050	0.907 ± 0.034	0.811
	Multimodal	Early Fusion	0.895 ± 0.037	0.880 ± 0.028	0.919
		Late Fusion	0.930 ± 0.032	0.877 ± 0.059	0.913
		Joint Fusion	0.981 ± 0.005	0.911 ± 0.022	0.883

The first two data columns show the results (mean and standard deviation (stdev)) on the cross-validation (CV) training and validation sets (10-fold CV for adults and fivefold CV for pediatric cohort). The last column shows the result of the final model on the test set.