Purpose Prediction of Torsades de Pointes (TdP) risk using hiPSC-CM assays remains challenging, as many models fail to capture nonlinear patterns and exhibit unstable performance across different drugs. We examined whether a stacking ensemble can improve robustness when only two simple <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$\Delta FPDc$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="normal">Δ</mi> <mi>F</mi> <mi>P</mi> <mi>D</mi> <mi>c</mi> </mrow> </math></EquationSource> </InlineEquation>-derived predictors are available. Methods Two electrophysiological predictors were derived from MEA-based FPDc measurements: the maximum change (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">$\Delta FPDc$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="normal">Δ</mi> <mi>F</mi> <mi>P</mi> <mi>D</mi> <mi>c</mi> </mrow> </math></EquationSource> </InlineEquation>) and the interpolated <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$\Delta FPDc$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi mathvariant="normal">Δ</mi> <mi>F</mi> <mi>P</mi> <mi>D</mi> <mi>c</mi> </mrow> </math></EquationSource> </InlineEquation> at Cmax. These features were used to train a stacking model comprising random forest (RF), XGB, and a shallow artificial neural network (ANN). Performance was assessed on 16 unseen CiPA reference compounds using AUC, likelihood ratios, pairwise accuracy, and classification error. Class weighting addressed class imbalance, and likelihood-based metrics assessed diagnostic consistency. Results The stacking ensemble consistently outperformed all single classifiers. The XGB-based meta-classifier achieved perfect discrimination for both AUC1 and AUC2 (1.000), with pairwise accuracy reaching 1.000 and classification error remaining below 0.125 across repeated evaluations. Conclusions Combining a simple MEA-derived feature set with a stacking architecture provides a more reliable framework for early TdP risk assessment than single-model approaches. External validation using additional MEA datasets and the integration of interpretable modeling strategies will be important for future translational use.

Stacking Ensemble Machine Learning for Cardiac Safety Assessment Using hiPSC-CM MEA Data

期刊论文
发表日期：

2026年1月26日

查看全文

Muhammad Adnan Pramudito,
Yunendah Nur Fuadah,
Yoo Seok Kim,
Ki Moo Lim

摘要

Purpose

Prediction of Torsades de Pointes (TdP) risk using hiPSC-CM assays remains challenging, as many models fail to capture nonlinear patterns and exhibit unstable performance across different drugs. We examined whether a stacking ensemble can improve robustness when only two simple $\Delta FPDc$ $Δ F P D c$ -derived predictors are available.

Methods

Two electrophysiological predictors were derived from MEA-based FPDc measurements: the maximum change ( $\Delta FPDc$ $Δ F P D c$ ) and the interpolated $\Delta FPDc$ $Δ F P D c$ at Cmax. These features were used to train a stacking model comprising random forest (RF), XGB, and a shallow artificial neural network (ANN). Performance was assessed on 16 unseen CiPA reference compounds using AUC, likelihood ratios, pairwise accuracy, and classification error. Class weighting addressed class imbalance, and likelihood-based metrics assessed diagnostic consistency.

Results

The stacking ensemble consistently outperformed all single classifiers. The XGB-based meta-classifier achieved perfect discrimination for both AUC1 and AUC2 (1.000), with pairwise accuracy reaching 1.000 and classification error remaining below 0.125 across repeated evaluations.

Conclusions

Combining a simple MEA-derived feature set with a stacking architecture provides a more reliable framework for early TdP risk assessment than single-model approaches. External validation using additional MEA datasets and the integration of interpretable modeling strategies will be important for future translational use.