a Identification of significant predictors for predictive model generation. Volcano plot showing mRNA and protein variables that are statistically significant for the prediction of BOR by unadjusted univariate analysis (p < 0.10, n = 527 variables, p < 0.05, n = 228 variables). b Combined modality model is superior to RNA-only or protein-only models in terms of BOR classification. Box and whisker plots and receiver operator characteristic (ROC) curves comparing a bulk RNA-only model (n = 770 variables; Area under the curve [AUC], 0.93; 95% confidence intervals [CI], 0.87–1.00; sensitivity, 0.93; specificity, 0.87; positive predictive value [PPV], 0.85; negative predictive value [NPV], 0.94) with a DSP-only model (n = 117 variables; AUC, 0.87; 95% CI, 0.80–0.94; sensitivity, 0.79; specificity, 0.88; PPV, 0.84; NPV, 0.84) and a combined bulk RNA and DSP model (n = 44 variables; AUC, 0.97; 95 CI, 0.92-1.00; sensitivity, 0.96; specificity, 0.93; PPV, 0.91; NPV, 0.96); feature selection occurs through Elastic Net Regularization after removal of moderately correlated predictors (R2 > 0.70). On each boxplot, the central line indicates the median and edges indicate the interquartile range. The upper whisker extends from the 75th percentile to the largest value at most the 1.5x interquartile and the lower whisker extends from the 25th percentile to the smallest value at most the x1.5 interquartile.