Long short-term memory recurrent neural network for pharmacokinetic-pharmacodynamic modeling

Area of Science:

• Pharmacometrics
• Computational Biology
• Machine Learning

Background:

• Recurrent neural networks (RNNs) are effective for time series analysis.
• Limited research exists on RNN applications in pharmacokinetic (PK) and pharmacodynamic (PD) analyses.

Purpose of the Study:

• To present and evaluate a Long Short-Term Memory (LSTM) network, a type of RNN, for PK/PD analysis.
• To analyze simulated PK/PD data of a hypothetical drug using an LSTM model.

Main Methods:

• Trained an LSTM model using plasma concentration and effect data from one dosing regimen.
• Utilized the trained LSTM model to predict individual PK/PD data under different dosing regimens.

Main Results:

• The optimized LSTM model successfully captured temporal dependencies in the data.
• Accurate prediction of PD profiles was achieved for a simulated indirect PK-PD relationship.

Conclusions:

• The developed LSTM model demonstrates the capability to approximate complex, underlying mechanistic biological processes.
• LSTM networks show promise as a tool for PK/PD analysis.