European Commission Framework Program 7 funded project, aims to develop distributed framework for predictive toxicology. The building blocks considered are : data, chemical structures, algorithms and models. The framework allows to build models, apply models, validate models, access and query data in various ways. Technologies used are REST style web services and W3C Resource Description Framework for description of services.
Application Programming Interface
  • The way applications talk to each other
  • The way developers talk to applications
OpenTox API
The Application Programming Interface of OpenTox framework
OpenTox components
OpenTox considers the following set of entities as essential building blocks of a predictive toxicology framework:
  • Chemical compound structure API Examples
    • Chemical compound properties and identifiers
  • Dataset of chemical compounds and various properties (measured or calculated) API Examples
  • Algorithms API Examples
    • Data processing algorithms
    • Algorithms generating certain values, based on chemical structure (e.g. descriptor calculation)
    • Data preprocessing (e.g. Principal component analysis, feature selection)
    • Structure processing (e.g. structure optimization)
    • Algorithms, relating set of structures to another set of structures (e.g. similarity search or metabolite generation)
    • Machine learning algorithms
      • Supervised (e.g. Regression, Classification)
      • Unsupervised (e.g. Clustering )
    • Prediction algorithms, defined by experts (e.g. series of structural alerts, defined by human experts , not derived by learning algorithms)
    • Applicability domain algorithms
  • Models API Examples

    Models are generated by applying machine learning algorithms to specific dataset , given specific parameters. In case of expert defined rules, models are generated by the corresponding algorithms, without requirement for a training dataset.

  • Validation API

    Validation provides procedures independent of model building facilities (e.g. crossvalidation) and generates relevant statistics.

  • Reports API

    Various types of reports might be generated, using building blocks listed (e.g. validation report can be generated using validation object, a model and a dataset).

In addition, the following components are introduced in order to address specific challenges in IT implementation, for example handling of time consuming calculations or access to protected resources

AMBIT API implementation and extensions

Back to top

Last Published: 2018-01-11.