Background

The goal of accessing ADME properties is to understand how a potential drug candidate interacts with the human body, including absorption, distribution, metabolism, and excretion. This knowledge is crucial for evaluating efficacy, safety, and clinical potential, guiding drug development for optimal therapeutic outcomes. Fang et al. 2023 has disclosed DMPK datasets collected over 20 months across six ADME in vitro endpoints, which are human and rat liver microsomal stability, MDR1-MDCK efflux ratio, solubility, and human and rat plasma protein binding. The dataset contains 885 to 3087 measures for the corresponding endpoints.

Benchmarking

The goal* of this benchmark is to perform a single task, which is to the best predictive model for human plasma protein binding.

Description of readout

• Readouts: LOG PLASMA PROTEIN BINDING (HUMAN) (% unbound)
• Bioassay readout: Human plasma protein binding
• Optimization objective: Lower value
• Number of data points: train: 2218 test: 559

Molecule data resource:

Reference: https://doi.org/10.1021/acs.jcim.3c00160

Train/test split:

In this benchmark set, the same train/test sets in the fang2023 paper were used for the 6 endpoints human and rat liver microsomal stability, MDR1-MDCK efflux ratio, solubility, and human and rat plasma protein binding, respectively. See more details at https://github.com/molecularinformatics/Computational-ADME/tree/main/MPNN.

Distribution of the train/test in the chemical space

Related links

The full curation and creation process is documented here.