Figure S1: Accuracy of Feature-Specific Classifiers from Deep-Learning Model for Tumor-Type Prediction Using Targeted Clinical Genomic Sequencing Data

For each broad category of features present within our training set, we trained individual models using the same training regime as GDD-ENS. Results are shown across these categories, represented by a circle. We then iteratively combined and retrained models, adding these broad feature groups in decreasing order of the accuracy of their individual model, represented by an X. The X corresponding to the category indicated on the X-axis corresponds to the accuracy of the model trained on all categories to the left of it. The model trained on all the features has the highest accuracy overall on held-out data. CN; Copy Number.