qInward variability-based in-silico proarrhythmic risk assessment of drugs using deep learning model

Figures

• 

  FIGURE 1. Schematic of the proposed algorithm for TdP-risk assessment; (A), flow chart of the process; (B), the proposed convolutional neural network (CNN) classifier using in silico feature variability; MCMC, Markov-chain Monte Carlo; H, Hill coefficients; IC50, the half inhibitory concentration; Conv1D, one-dimensional CNN layer; Batch Norm, Batch Normalization; MaxP 1D, one-dimensional max pooling layer; str, strides; ReLU, Rectified Linear Unit activation function.

• 

  FIGURE 2. Schematic of 10,000-test algorithm; CNN, convolutional neural network model; AUC, area under the receiver operating curve; LR, likelihood ratio.

• 

  FIGURE 3. Distribution of AUCs based on the TdP-risk using qInward variability of 16 test drugs in the Chantest dataset; (A–C), AUC distribution for the high, intermediate, and low-risk of the CNN classifier for 16 test drugs.

• 

  FIGURE 4. Distribution of AUCs based on the TdP-risk using qInward variability merged of three datasets; (A–C), AUC distribution for the high, intermediate, and low-risk of the CNN classifier for 16 test drugs; (D–F), AUC distribution for the high, intermediate, and low-risk for the CNN classifier for 28 drugs.