Fig. 8
From: Systemic evolutionary chemical space exploration for drug discovery
Test data \(R^{2}\) of deep learning models. All the docking data are randomly split into three parts: training (80%), test (10%), and validation (10%) datasets. A–E) shows the performance of models using docking data set from generation 1–5, respectively. F shows the performance of the aggregate dataset, including data from the previous five generations