model,auroc,pr_auc,ef1,ef5,n_actives,n_decoys,scaffold_split,set,formula
AI screen / complementarity (full library),0.963,0.502,50.0,10.0,4,8496,True,Y220C published ligands vs property-matched library decoys,AUROC = Mann–Whitney P(score_active > score_decoy); EF = (hits in top x%) / expected
Docking −ΔG (shortlist + controls),0.5,0.507,55.0,10.0,4,216,True,Y220C published ligands vs property-matched library decoys,AUROC = Mann–Whitney P(score_active > score_decoy); EF = (hits in top x%) / expected
Mutant selectivity (shortlist + controls),0.243,0.012,0.0,0.0,4,216,True,Y220C published ligands vs property-matched library decoys,AUROC = Mann–Whitney P(score_active > score_decoy); EF = (hits in top x%) / expected
MD occupancy (shortlist + controls),0.497,0.338,27.5,10.0,4,216,True,Y220C published ligands vs property-matched library decoys,AUROC = Mann–Whitney P(score_active > score_decoy); EF = (hits in top x%) / expected
Full Rescue score (shortlist + controls),0.731,0.513,55.0,10.0,4,216,True,Y220C published ligands vs property-matched library decoys,AUROC = Mann–Whitney P(score_active > score_decoy); EF = (hits in top x%) / expected