Fig 1.
Membrane depolarization of spontaneously beating human induced pluripotent stem cell-derived cardiomyocytes (hiPS-CM).
Cells were stained with voltage sensitive dye, di-4-ANE(F)PPTEA (A). Corresponding brightfield image (B). Consecutive line scans were performed using 2-photon microscopy on sets of individual cell membranes as shown in (A) and the corresponding zoom image (C). The resultant temporal sequence of line scans with heat-mapped fluorescence intensity prior to motion correction (M 1–4 specifies the individual membrane numbers) (D). Subsequent sample quantification of fluorescence intensity of a given membrane over time (M3 is the subset of the temporal line scan sequence for membrane 3) (E).
Fig 2.
Drug induced alteration of membrane depolarization waveforms of spontaneously beating hiPS-CM.
Representative traces (corrected for photobleaching) are presented for the control, N = 145 (A), propranolol-treated, N = 100 (B), and isoproterenol-treated, N = 216 (C) were administered to separate cultures. Quantification of beat rate (D). All error bars are SE. (*) indicates p<0.05 for a given comparison.
Fig 3.
Metric definitions: The maximum height (hmax) is defined as the maximum amplitude of the waveform. The upslope (mup) and downslope (mdown) are defined as the slopes at 50% of the maximum height. The peak width (w) is defined as the distance from upslope to downslope at 50% of the maximum height. The plateau height (hplateau) is defined as the height of the waveform at the midway point of the peak width (A). Average of waveform metrics (B). All error bars are SE. (*) indicates p<0.05 for a given comparison.
Fig 4.
Metrics with the least relative importance to classification were removed sequentially. Blue = 5 parameter model, Red = 3 parameter model (upslope (mup) and downslope (mdown) removed), Green = 2 parameter model (upslope (mup), downslope (mdown), and width (w) removed). Relative classification errors as a function of number of grown decision trees (A). Relative importance of waveform metrics to classification determination (B). Accuracy of classification using training data set in matrix format (C) and FPR/TPR plots (D). AUC(N) = area under curve for ‘N’ parameter model.
Table 1.
Model Accuracy as a Function of the Number of Model Parameters.
Fig 5.
Randomized blinded “unknown” waveform data classification.
Performed using reduced models (blue shades = 5 parameter model, red shades = 3 parameter model, green shades = 2 parameter model). Data are mean probability of correct classification of a single waveform as each treatment +/- SE (Con = control, Iso = isoproterenol, Pro = propranolol). The x-axis defines the actual treatment of a given waveform. (*) indicates p<0.05 for a given comparison between models (number of parameters) within a predicted treatment group.
Table 2.
Model Accuracy by Treatment as a Function of the Number of Model Parameters.