McKenna E. Buckley

Oxazolidinones are a broad-spectrum class of synthetic antibiotics that bind to the 50S ribosomal subunit of Gram-positive and Gram-negative bacteria. Many crystal structures of the ribosomes with oxazolidinone ligands have been reported in the literature, facilitating structure-based design using methods such as molecular docking. It would be of great interest to know in advance how well docking methods can reproduce the correct ligand binding modes and rank these correctly. We examined the performance of five molecular docking programs (AutoDock 4, AutoDock Vina, DOCK 6, rDock, and RLDock) for their ability to model ribosomal–ligand interactions with oxazolidinones. Eleven ribosomal crystal structures with oxazolidinones as the ligands were docked. The accuracy was evaluated by calculating the docked complexes’ root-mean-square deviation (RMSD) and the program’s internal scoring function. The rankings for each program based on the median RMSD between the native and predicted were DOCK 6 > AD4 > Vina > RDOCK >> RLDOCK. Results demonstrate that the top-performing program, DOCK 6, could accurately replicate the ligand binding in only four of the eleven ribosomes due to the poor electron density of said ribosomal structures. In this study, we have further benchmarked the performance of the DOCK 6 docking algorithm and scoring in improving virtual screening (VS) enrichment using the dataset of 285 oxazolidinone derivatives against oxazolidinone binding sites in the S. aureus ribosome. However, there was no clear trend between the structure and activity of the oxazolidinones in VS. Overall, the docking performance indicates that the RNA pocket’s high flexibility does not allow for accurate docking prediction, highlighting the need to validate VS. protocols for ligand-RNA before future use. Later, we developed a re-scoring method incorporating absolute docking scores and molecular descriptors, and the results indicate that the descriptors greatly improve the correlation of docking scores and pMIC values. Morgan fingerprint analysis was also used, suggesting that DOCK 6 underpredicted molecules with tail modifications with acetamide, n-methylacetamide, or n-ethylacetamide and over-predicted molecule derivatives with methylamino bits. Alternatively, a ligand-based approach similar to a field template was taken, indicating that each derivative’s tail groups have strong positive and negative electrostatic potential contributing to microbial activity. These results indicate that one should perform VS. campaigns of ribosomal antibiotics with care and that more comprehensive strategies, including molecular dynamics simulations and relative free energy calculations, might be necessary in conjunction with VS. and docking.

A promising protein target for computational drug development, the human cluster of differentiation 38 (CD38), plays a crucial role in many physiological and pathological processes, primarily through the upstream regulation of factors that control cytoplasmic Ca²⁺ concentrations. Recently, a small-molecule inhibitor of CD38 was shown to slow down pathways relating to aging and DNA damage. We examined the performance of seven docking programs for their ability to model protein-ligand interactions with CD38. A test set of twelve CD38 crystal structures, containing crystallized biologically relevant substrates, were used to assess pose prediction. The rankings for each program based on the median RMSD between the native and predicted were Vina, AD4 > PLANTS, Gold, Glide, Molegro > rDock. Forty-two compounds with known affinities were docked to assess the accuracy of the programs at affinity/ranking predictions. The rankings based on scoring power were: Vina, PLANTS > Glide, Gold > Molegro >> AutoDock 4 >> rDock. Out of the top four performing programs, Glide had the only scoring function that did not appear to show bias towards overpredicting the affinity of the ligand-based on its size. Factors that affect the reliability of pose prediction and scoring are discussed. General limitations and known biases of scoring functions are examined, aided in part by using molecular fingerprints and Random Forest classifiers. This machine learning approach may be used to systematically diagnose molecular features that are correlated with poor scoring accuracy.