Fig. 4: Flexynesis can be trained concurrently for all three types of tasks: regression, classification, and survival at a single run.

The model was trained on 557 training samples from the merged cohort of the LGG (Lower Grade Glioma) and GBM (Glioblastoma Multiforme) patient samples with three supervisor heads: a regressor for the patient age (AGE), a classifier for the histological diagnosis, and a survival head for the overall survival status of the patient (OS_STATUS). A Displays the tSNE (t-distributed Stochastic Neighbor Embedding) visualization of the sample embeddings for 239 test samples, where the size of the points reflect the age of the patient, the colors represent the histological diagnosis, and the samples were stratified into high-risk and low-risk groups based on the predicted risk scores for each patient. The sample embeddings reflect the impact of all three clinical variables concurrently. B Displays the top 10 most important features discovered for each supervisor head for the patient’s age, histological subtype, and survival status. Source data are provided as a Source Data file.