Drug‐induced shortening of the electromechanical window is an effective biomarker for in silico prediction of clinical risk of arrhythmias

Figures

• 

  Figure 1. Drug‐induced AP phenotypes and APD90, CTD90 and EMw biomarkers. (a,b) Representative AP traces illustrating repolarisation abnormalities (a) and depolarisation abnormalities (b). (c) In silico EMw, defined here as the difference between CTD90 and APD90

• 

  Figure 2. Optimised population of human ventricular models. (a) AP and Ca2+ transient traces of the 107 models in the calibrated population (blue and green respectively). (b) Ionic profile of the calibrated population, shown as scaling factors with respect to the baseline model. Central mark corresponds to median, box limits are the 25th and 75th percentiles, and whiskers extend to the most extreme data points. (c) AP and Ca2+ transient biomarker distributions for the calibrated population (blue histograms) and experimental ranges (black lines): AP duration at 40%, 50%, and 90% of repolarisation (APD40, APD50, and APD90); AP triangulation, as difference between APD90 and APD40 (Tri90‐40); Ca2+ transient duration at 50% and 90% of decay (CTD50 and CTD90)

• 

  Figure 3. Drug‐induced changes in EMw (∆EMw) in the in silico population for 40 compounds at 10× EFTPCmax. Data are shown using the log‐modulus transformation, as described in Section 2. ∆EMw shortening threshold (−10%) is marked as a dashed red line. Colour scale (green to red) corresponds to TdP risk category: no TdP risk (green), conditional (yellow), possible (orange), and known (red) TdP risk. Drug‐induced RA occurrence is marked by one or more asterisks, based on the fraction of models displaying RA: * (>10%), ** (>30%), *** (>50%). Boxplots are not shown for drugs exhibiting more than 97% RA (three models or less for which to compute the EMw), and these are depicted on the dotted line at the bottom of the figure. Boxplots description as in Figure 2. Result obtained by using the single ORd model are shown as black dots, connected by a dashed black line

• 

  Figure 4. Clinical TdP risk predictions for 40 reference compounds, at five concentrations (multiples of EFTPCmax) in the human in silico population of models, using RA occurrence only (top line, in blue), or using RA + ∆EMw (bottom line, in orange). Predictions with accuracy >80% are shown in bold. Predictions with highest accuracy are highlighted with a rectangle

• 

  Figure 5. TdP scores, computed for the 40 reference compounds considering RA only (a) and RA + ∆EMw (b). Compounds with a low TdP score are predicted to be safe, while compounds with a high TdP score are predicted to be risky. Drug TdP risk categories as in Figure 3, from green to red

• 

  Figure 6. Relationship between APD90, CTD90, and EMw, in the 107 human ventricular AP models in control. All data are colour coded (bar on the right) based on the magnitude of the corresponding EMw, from 0 ms (yellow) to 410 ms (dark pink). (a) AP traces for the control population. (b) EMw versus APD90, and their regression line. (c) Ca2+ transient traces for the control population. (d) EMw versus CTD90, and their regression line

• 

  Figure 7. (a) Effect of multiple ion channel blocks (IKr, ICaL, and INaL) on APD90 (bottom row) and EMw (top row). Each column corresponds to a fixed percentage of INaL block (from 0% to 100%, left to right), and each single plot summarises the results for the 25 combinations of IKr and ICaL blocks (y and x axis respectively). Values shown are the median changes in the 107 human models. “‐“ indicates that value could not be obtained due to occurrence of RA or lack of Ca2+ transient (e.g., total ICaL block). Opposite colour scales were used for EMw and APD90, from white to black for EMw and from black to white for APD90. Since drug‐induced TdP risk is associated with EMw shortening and APD90 prolongation, for both biomarkers, white is therefore associated with high risk. (b) Effect of multiple ion channel blocks on the fraction of models displaying RA occurrence (top row), EMw shortening beyond threshold (middle row), and APD90 prolongation beyond threshold (bottom row). Numerical values are all in the range (0–100)%, coloured from black to white, and higher values correspond to block combinations predicted at higher risk for that specific biomarker