Fig. 4: The effect of additional docking constraints on a subset of three crystal structures 4EIY, 5OLH and 6GT3.

a Number of unique chemotypes identified according to docking constraints used. b Total number of unique molecules generated that possess known A_2A chemotypes according to docking constraints used. c Proportion of molecules containing A_2A chemotypes per step during optimisation according to docking constraints used. Error bands (a–c) show the 95% confidence interval. d Additional A_2A chemotypes identified with the use of stricter docking constraints. e Binding site representation of the crystal structure 6GT3. The receptor is shown in grey cartoon with residue side chains in wireframe representation with grey carbons, whereas N253^6.55 and co-crystallised ligand in stick representation, with the ligand carbon atoms coloured in purple. Hydrogen bonds to N253^6.55 are shown with grey dashed lines. GRID maps are shown as transparent solids, with the pocket surface, in terms of how close a ligand carbon atom can approach, contoured by a CH3 methyl probe at 1 kcal/mol (in grey) and lipophilic hotspots sub-pockets contoured by a sp2 CH probe (C1=) probe at −2.8 kcal/mol (in yellow). The lipophilic hot spots annotated in sub-pockets I to III are respectively occupied by the triazine, the 4-fluorophenyl and the substituted pyridyl rings of the ligand; sub-pocket IV contains another lipophilic hot spot located above pocket III. WaterFLAP water networks calculated on the pseudo-apo binding site (with the ligand removed) are shown as spheres and colour-coded by relative energetic scoring in red when predicted free energy (ΔG) is higher than 3.0 kcal/mol, yellow when ΔG is between 1.5 and 3.0 kcal/mol, grey if ΔG is between ‒2.0 and 1.5 kcal/mol, and blue when ΔG <‒2.0 kcal/mol. Source data are provided as a Source Data file.