Improving the hERG model fitting using a deep learning-based method.

Area of Science:

• Computational Biology
• Pharmacology
• Machine Learning

Background:

• The hERG channel is vital for cardiac action potential and drug toxicity assessment.
• The Comprehensive in vitro Proarrhythmia Assay (CiPA) protocol faces challenges in kinetic effect identification.
• Model-based parameter identification for hERG channel kinetics is time-consuming and prone to local minima issues.

Purpose of the Study:

• To propose a deep learning-based method for improving model fitting of drug effects on the hERG channel.
• To address the challenges of time-consuming parameter inference and fitting failures in hERG channel modeling.
• To provide appropriate initial values for model parameter identification.

Main Methods:

• Generated a dataset by altering model parameters and training a deep learning model.
• Utilized spectrograms incorporating time, frequency, and amplitude to enhance accuracy.
• Trained the deep learning model on simulated hERG model data and validated it with experimental data.

Main Results:

• Successfully identified appropriate initial values for model parameter fitting.
• Significantly improved the speed of parameter fitting.
• Successfully avoided fitting failures caused by local minima.

Conclusions:

• The proposed deep learning method effectively improves model fitting for hERG channel kinetics.
• This approach accelerates drug development and toxicity identification by reducing analysis time and effort.
• The method is applicable to various in silico models for diverse applications beyond cardiac safety.