Figure 2

Training data for the 6-mRNA classifier. (a) Visualization of 705 samples across 21 studies in low dimension using t-SNE. (b) Logistic regression model selection. Each dot corresponds to a model defined by a combination of logistic regression hyperparameters and a decision threshold. Entire search space (100 hyperparameter configurations) is shown. (c) ROC plot for the best model. The plot is constructed using pooled probabilities from cross-validation folds. (d) Expression of the 6 genes used in the logistic regression model according to mortality outcomes.