Welcome to Admet-XSpec’s documentation!

ADMET-XSpec is an open-source tool that facilitates systematic cross-species data integration for training ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) prediction models. The tool is designed to help researchers assess and understand the translatability of ADMET parameters measured in non-human assays in the context of human drug development.

ADMET-XSpec implements a comprehensive workflow that includes data preprocessing, feature engineering, model training, and evaluation. ADMET-XSpec allows for the evaluation of predictive models trained on human data augmented by non-human data points, leveraging a selection of machine learning algorithms, feature extraction techniques and additional data preprocessing steps.

Our source code is hosted on Github: the ADMET-XSpec repository

Note

This project is under active development. We expect to increment and restructure over the following months, and cannot guarantee the stability of all interfaces.

Contents:

• Overview
  • Three experiments you can run right now
  • Killer features
• Quick start
  • Installation with conda
  • Installation with uv
  • Running classical ML experiments
  • Running deep ML experiments
• Guides
  • Choosing your “processing plan”
  • Data exploration and visualization
  • Sourcing and setting up data
  • Training and hyperparameter optimization
• API
  • src.data.data_interface
  • src.data.featurizer
  • src.processing_pipeline