Predicted Mode of Binding to and Allosteric Modulation of the µ-Opioid Receptor by Kratom's Alkaloids with Reported Antinociception In Vivo.

Pain management devoid of serious opioid adverse effects is still far from reach despite vigorous research and development efforts. Alternatives to classical opioids have been sought for years, and mounting reports of individuals finding pain relief with kratom have recently intensified research on this natural product. Although the composition of kratom is complex, the pharmacological characterization of its most abundant alkaloids has drawn attention to three molecules in particular, owing to their demonstrated antinociceptive activity and limited side effects in vivo. These three molecules are mitragynine (MG), its oxidized active metabolite, 7-hydroxymitragynine (7OH), and the indole-to-spiropseudoindoxy rearrangement product of MG known as mitragynine pseudoindoxyl (MP). Although these three alkaloids have been shown to preferentially activate the G protein signaling pathway by binding and allosterically modulating the µ-opioid receptor (MOP), a molecular level understanding of this process is lacking and yet important for the design of improved therapeutics. The molecular dynamics study and experimental validation reported here provide an atomic level description of how MG, 7OH, and MP bind and allosterically modulate the MOP, which can eventually guide structure-based drug design of improved therapeutics.

Hidden bias in the DUD-E dataset leads to misleading performance of deep learning in structure-based virtual screening.

Recently much effort has been invested in using convolutional neural network (CNN) models trained on 3D structural images of protein-ligand complexes to distinguish binding from non-binding ligands for virtual screening. However, the dearth of reliable protein-ligand x-ray structures and binding affinity data has required the use of constructed datasets for the training and evaluation of CNN molecular recognition models. Here, we outline various sources of bias in one such widely-used dataset, the Directory of Useful Decoys: Enhanced (DUD-E). We have constructed and performed tests to investigate whether CNN models developed using DUD-E are properly learning the underlying physics of molecular recognition, as intended, or are instead learning biases inherent in the dataset itself. We find that superior enrichment efficiency in CNN models can be attributed to the analogue and decoy bias hidden in the DUD-E dataset rather than successful generalization of the pattern of protein-ligand interactions. Comparing additional deep learning models trained on PDBbind datasets, we found that their enrichment performances using DUD-E are not superior to the performance of the docking program AutoDock Vina. Together, these results suggest that biases that could be present in constructed datasets should be thoroughly evaluated before applying them to machine learning based methodology development.