Fig. 3

The SVM classifier used for evaluating TASA. A Schematic summary of the classification scheme. Training and test samples were obtained from GEO on CD8, Breast, and Monocyte datasets. Training was conducted on the first 10 PCs of the training dataset (random 80% of samples) and testing was conducted on the test dataset (remaining 20%). Next, the trained model was used to classify simulated samples. B The average of absolute decision values from the SVM classifier is used as an evaluation score for each simulation approach. Each simulation approach is color-coded according to its score. The higher the score, the better the simulation. C PCA plot of the simulated and real datasets. As the best performer of all approaches, S4 generated the simulated samples shown in this panel. Tissue attributes are represented by the shape of dots, and batches with colors (red for simulated data, green for test data, and blue for train data)