New synthetic cannabinoids and the potential for cardiac arrhythmia risk

Figures

• 

  Fig. 1. Schematic diagram illustrating prediction approaches used. Between December 2022 and April 2023, SMILES (Simplified Molecular Input Line Entry System) structures (such as that shown for 5F-BZO-POXIZID) for each compound evaluated were entered into the online user interface of the two evaluation tools. For compounds within its applicability domain (AD) PredhERG [13] provided binary (blocker/non-blocker) and multiclass (non-blocker, weak-moderate or strong blocker) predictions (see Table 1). The reported correct classification rates are 0.83–0.84 for the binary model and an accuracy of 0.66–0.79 for the multiclass model [13]. PredhERG also provides a visual output (example shown for 5F-BZO-POXIZID) highlighting in green molecular fragments recognized by the software that are likely to contribute to hERG block [13]. hERGSPred does not provide AD information and provides a non-synonymous output to PredhERG, evaluating the probability of a compound being a hERG blocker (output values between 0 and 1; see Table 1) using a consensus model with an accuracy of 0.839 [14].

• 

  Fig. 2. Illustrative prediction of action potential (AP) prolongation for different potency hERG blocking compounds. Ventricular AP simulations were conducted in August 2023 using the “AP Predict” online cardiac electrophysiology simulator [18]. From within the simulator the O'Hara Rudy CiPA ventricular AP model was selected and pIC50 values corresponding to strong/extreme (6.523), weak/moderate (5.523) and non-blocker (4.523) predicted inhibitory ranges of PredhERG were entered. APs were elicited at 1 Hz and for each pIC50 effects on APD90 of 0.1, 0.3, 1 and 3 μM were evaluated. (A) shows APs in control and at each inhibitor concentration, for a pIC50 of 6.523 (IC50 of 0.3 μM). (B) shows a plot of extent of prolongation of APD90 (ΔAPD90 (%)) at each test concentration for the 3 simulated different potency inhibitors.