qFit-ligand algorithm workflow. All ligands undergo three preliminary searches: unconstrained, fixed terminal atoms, and blob search, allowing varying degrees of freedom (A-C). If the ligand has short or long side chains, the algorithm progresses to more specialized searches: branch search for ligands with side chains of at least four atoms (D), and long chain search for those exceeding 30 atoms (E). The algorithm then determines the best fit of generated conformers to electron density through quadratic programming, followed by additional sampling with rotations and translations (F). The remaining conformers then undergo quadratic and mixed-integer quadratic programming to ensure that only the most well-supported conformers are included in the final model.

Analysis of ligand conformations generated by qFit-ligand. (A) Differences in RSCC (x-axis) and torsion strain (y-axis) between qFit-ligand predicted structures and modified true positives. The lower right quadrant shows structures for which we improve both RSCC and strain. (B) Gallery of examples for which the updated qFit-ligand models have improved RSCC and strain compared to the modified true positives. The composite omit density map is contoured at 1σ for every structure. (C) Differences in RSCC and torsion strain between the updated qFit-ligand and the original qFit-ligand. Figure 3. (A) RSCC of the synthetic true benchmark structures plotted against map resolution (in Ångstroms) for different conformer occupancy ratios, showing a decrease in RSCC with deteriorating map resolution. (B) RSCC of qFit-ligand generated multiconformer models, plotted against map resolution and grouped by conformer occupancy split. (C) RMSD between the closest qFit-ligand conformer and the true ‘B’ conformer. (D, left) True structure and qFit-ligand predicted structure of 3SC multiconformer ligand with a map resolution of 0.8 Å and conformer occupancy split of 0.50/0.50. (D, right) True structure and qFit-ligand predicted structure of 3SC multiconformer ligand with a map resolution of 0.8 Å and conformer occupancy split of 0.80/0.20. The lower right quadrant shows structures for which we improve both RSCC and strain.

(A) RSCC of the synthetic true benchmark structures plotted against map resolution (in Ångstroms) for different conformer occupancy ratios, showing a decrease in RSCC with deteriorating map resolution. (B) RSCC of qFit-ligand generated multiconformer models, plotted against map resolution and grouped by conformer occupancy split. (C) RMSD between the closest qFit-ligand conformer and the true ‘B’ conformer. (D, left) True structure and qFit-ligand predicted structure of 3SC multiconformer ligand with a map resolution of 0.8 Å and conformer occupancy split of 0.50/0.50. (D, right) True structure and qFit-ligand predicted structure of 3SC multiconformer ligand with a map resolution of 0.8 Å and conformer occupancy split of 0.80/0.20.

Analysis of ligand conformations generated by qFit-ligand on the un-biased modified true positive dataset. (A) Distribution of the number of conformers output by qFit-ligand. (B) Differences in RSCC and torsion strain between the qFit-ligand and the modified true positives. The lower right quadrant shows structures for which we improve both RSCC and strain.

Evaluation of qFit-ligand predicted macrocycle conformations. (A) Differences in RSCC and torsion strain between qFit-ligand predicted structures and refined deposited single conformer macrocycles. The lower right quadrant shows structures for which we improve both RSCC and strain. (B) Gallery of examples for which the qFit-ligand models have improved RSCC and strain compared to the deposited single conformer macrocycle ligand. The composite omit density map is contoured at 1σ for every structure.

(A) RMSD between the deposited ‘B’ conformer and the closest qFit-ligand conformer. Lower values correlate with a closer recapitulation of the deposited heterogeneity. (B) RSCC and torsion strain differences in the deposited models and the qFit-ligand predicted models. The lower right quadrant shows structures for which we improve both RSCC and strain. (C) Gallery of examples for which qFit-ligand successfully recovers well-fitting alternate conformers. The composite omit density map is contoured at 1σ for every fragment.

RDKit determines a distance bounds matrix for a molecule by establishing upper and lower bounds for interatomic distances. These bounds are informed by experimental data and chemical knowledge of bond length, angle, and dihedral angle preferences obtained from the Cambridge Structural Database. Within a torsion angle formed by four atoms, the minimum distance between atoms 1 and 4 corresponds to the syn conformation, and the maximum distance corresponds to the anti conformation. These specific distances, d for syn and d’ for anti, are recorded in the bounds matrix as the lower and upper bounds, respectively. This is performed for every distance between each atom in the molecule. Randomly sampling these bounds with RDKit’s implementation of ETKDG gives rise to different conformations of the torsion angle.

Correlation between the number of atoms in the input ligand and total qFit-ligand runtime. A strong Pearson correlation of 0.66 indicates that as you increase the size of your input molecule, qFit-ligand will take longer to run.

Construction of the development true positive dataset and the unbiased true positive dataset.

Original (unmodified) multiconformer true positives compared to qFit-ligand conformers. The deposited ‘A’ conformer is shown in gray and the deposited ‘B’ conformer in green. The qFit-ligand conformer closest to the deposited ‘B’ is shown in purple. This demonstrates qFit-ligand’s ability to accurately recapitulate the original deposited multiconformer model. The composite omit density map is contoured at 1σ for every structure.

Comparison of torsion strain between qFit-ligand models before and after refinement, as well as the deposited structures. The five outlier structures where the refined qFit-ligand model strain exceeded the deposited model strain by more than 1 kcal/mol are highlighted.

Performance comparison of new and old qFit-ligand algorithms. (A) RSCC of new versus old qFit-ligand predicted conformations across the true positive dataset. Points above the diagonal line are for structures where the new qFit-ligand model has a higher (better) RSCC. (B) Torsion strain of new versus old qFit-ligand predicted conformations across the true positive dataset. Bars to the left of the vertical line are for structures where the new qFit-ligand model has improved (lower) internal strain. (C) Gallery of examples for which the updated qFit-ligand models are both higher in RSCC and lower in strain compared to the old qFit-ligand models. The composite omit density map is contoured at 1σ for every structure.

Modified true positive dataset comparison of new versus old qFit-ligand outlier cases. Modified true positive model (input for qFit-ligand), new qFit-ligand model, and old qFit-ligand model for PDB 2JJK, showing their respective RSCCs, strain, and conformer occupancies. This highlights significant improvements in both RSCC and strain. The composite omit density map is contoured at 1σ.

The four ligand multiconformer models from which our synthetic dataset was built. Here, they are shown at a map resolution of 0.8 Å at one sigma.

(A) Distribution of the number of conformers in qFit-ligand output models, showing varied conformer presence with a median of two conformers per structure. (B) Correlation between the number of conformers output by qFit-ligand and the RSCC of the input model. Higher input RSCC tends to yield a lower number of qFit-ligand conformers. (C) Comparison of strain between the single conformer deposited macrocycle and the qFit-ligand ‘B’ conformer for PDB 4Z2G using the COOT ligand distortion tool. The penalty scores for the two most distorted bonds and angles in the deposited model (left), compared to the same bonds and angles in the qFit-ligand ‘B’ conformer (right), demonstrating reduced strain in the alternate conformation.

The two structures for which qFit-ligand decreases RSCC and increases torsional strain. The composite omit density map is contoured at 1σ for both structures.