Fig. 2: Comparison of high-energy atom detector (HEAD) and PoseBusters⁸ in assessing ligand conformation validity.

a Valid conformation recall rates of HEAD and PoseBusters tested using Cambridge Structural Database (CSD) and LigBoundConf. Molecular conformations in CSD and LigBoundConf are all considered as valid, representing valid conformations without and with protein binding, respectively. All bar chart values are obtained from a single deterministic processing of the dataset; repeated runs yield identical results. b Histogram distribution of molecules in GM-5K dataset across ΔE with the bin size of 20 kcal/mol, where \(\Delta {{{\rm{E}}}}={{{{\rm{E}}}}}_{{{{\rm{ori}}}}}^{{{{\rm{DFT}}}}}-{{{{\rm{E}}}}}_{{{{\rm{opt}}}}}^{{{{\rm{DFT}}}}}.\) The red dashed line indicates the threshold value used to classify molecular conformations as valid or invalid. Molecules with ΔE values below the threshold are labeled as valid, while those above the threshold are labeled as invalid. The dynamic adjustment of the threshold value leads to corresponding changes in valid and invalid labels. c Weighted F1 scores of HEAD and PoseBusters across varying labeling thresholds based on ΔE in the GM-5K dataset. The threshold was systematically varied from 0 to 1000 kcal/mol in increments of 1 kcal/mol, resulting in a series of weighted F1 scores. The weighted F1 score is computed by first evaluating the F1 scores for both valid and invalid classes. Subsequently, these F1 scores are averaged considering the number of samples in each class as weights. Panels (d–g) showcase four representative abnormal geometries with their atomic energies, including steric clash (d), twisted ring structure (e), misplacement of a hydrogen atom (f), and valence violation (g). In panel (d), the steric clash between the chlorine atom and a methylene group is circled in red. In each panel, the molecular 3D conformation is shown as sticks, accompanied by a bar chart displaying atomic energies. Atom names are shown on the horizontal axis and energy values on the vertical axis. Atoms exhibiting abnormally high atomic energies are indicated by red arrows in the stick representation and highlighted by red bars in the corresponding bar charts.