Figure 5 Discussion

[gwaybio]

Figure 5. Predicting compound mechanisms of action (MOA) in Cell Painting and L1000 reveals complimentary performance for different mechanisms. (a) Test set model performance metrics for the multi-label, multi-class prediction framework. We trained models from a recent Kaggle competition plus a K nearest neighbors baseline model. The dotted bar chart represents a negative control in which we trained models with shuffled labels. The solid lines indicate ensemble model performance by blending model predictions (see methods). (b) Individual MOA performance by area under the precision-recall curve (AUPR) in the top performing model using Cell Painting and L1000 data. The dotted grey lines indicate a negative control baseline. We report all performance metrics using test set profiles.