Deep learning for pharmacokinetics and pharmacodynamics

Analysis of pharmacokinetics (PK) and pharmacodynamics (PD) through conventional methods such as non-linear mixed effect model, is complex and time-consuming to tune manually, and is therefore not compatible with the R&D speed needed in pharmaceutical industry. Recent publications showed that deep learning-based methods are promising as they significantly reduced the time and knowledge needed to build PKPD models while keeping high performance.

The study aims to 1) develop and evaluate deep learning methods for PKPD analysis; 2) pressure test and benchmark deep learning methods with simulations of PKPD under different scenarios such as noise, missing data etc.