Comparing the predicted binding affinity of a variant to that of the wildtype provides a relative measure of the variant impact on drug binding (Fig. 2). Since lower energies correspond to greater thermodynamic stability, increases in the relative binding energy correlate to decreased binding affinity of the drug compared to the wildtype receptor simulations. By calculating the relative binding energy and normalizing by the binding affinity to the wild-type receptor, we can more fairly compare ligand-protein binding energy across ligands. The full simulation results can be found in Table S1, along with code to execute their visualization using an interactive R Shiny app.
In general, variants show a similar impact across the opioid library, enhancing (238I, 302I) or diminishing (235M, 235N) the interactions of OPRM1 with the ligand. However, depending on the ligand, the 153V variant predicted both diminished (methadone, fentanyl) and enhanced binding (oxycodone, oliceridine) compared to the wildtype structure. The 153V variant in particular showed larger magnitude shifts (with a smaller range) in binding energy for multiple ligands, compared to the relatively large variance in results for ligands and other variants.
The natural neurotransmitter endomorphin-1 did not show drastic change in binding energy across variant receptors, which seems reasonable given these variants are not associated with any adverse conditions and found in healthy individuals. Several other ligands display a similar pattern of robustness (Fig. 3) across variant receptors. These ligands serve as possible candidates for widely prescribable pain treatment due to their robustness to patient genotype.
Figure 4 shows the binding profiles for the naturally occurring endomorphin-1 and morphine, and the widely used synthetic drugs fentanyl and oxycodone. Endomorphin 1 exhibits a higher variance in binding energy across trials compared to the remaining 3 drugs. Fentanyl, morphine, and oxycodone all showed relatively decreased binding energy (higher binding affinity) in variants 238I and 302I, suggesting that lower dosages may achieve the same analgesic effect (while minimizing risk of overdependence). The 153V variant varies significantly among these drugs, suggesting that a more curated approach to drug prescription may be needed for patients with this particular missense variant.
The drug naloxone is used widely to treat opioid overdose, and works by acting as a competitive antagonist for the opioid receptor. The drug is also often used in combination with other opioids to minimize the risk of overdependence, though the correct dosage in these situations still contains uncertainty (21). The docking simulation results indicate that naloxone may be robust to variants in the mu opioid receptor, though the relative binding energy decreased slightly in the 302I variant. However, no variant showed an increase in relative binding energy, indicating that individualized dosage might not be necessary when administering naloxone to prevent an overdose, which is often done in emergency situations where individual genotype information is not available.
The docking simulation results for each drug to the wild-type receptor were compared to experimentally determined Ki inhibitory constants (20), and an exponential regression model (Fig. 5) was built to predict the magnitude of change in Ki given the predicted change in binding energy for a drug-variant pair, relative to the wildtype receptor binding energy. Figure 6 shows the predicted range of Ki values for each variant-drug combination, binned into three ranges: ≤ 1 nM, 1-100 nM, and > 100 nM. Similar patterns of enhancing and diminishing effects can be seen across variants, though most variants were robust to extreme changes in predicted Ki, relative to the wild-type. In certain drugs, multiple variants increase the Ki by an order of magnitude (hydromorphone, buprenorphine, oxymorphine, alfentanil), while in other drugs, variants have the opposite effect (codeine, propoxyphene, fentanyl). Of particular note, for the drug pentazocine, certain variants decreased the Ki by two orders of magnitude, indicating a highly significant change in drug binding affinity, that could affect the potency of the drug and require a lower overall dosage in patients with those specific variants. Though the model excluded pentazocine as an outlier, the docking simulation results independently predict a trend of increasing the sensitivity for all variant receptors to the drug. The outlying Ki of pentazocine may also be explained by experimental variability and the fact that it is only a partial agonist for the mu opioid receptor (a characteristic which is not captured by simulated docking experiments) (20). The experimental error in determining inhibitory constants highlight the importance of more stringent validation experimental evidence for a clinically acceptable model.