Open Source Bayesian Models. 1. Application to ADME/Tox and Drug Discovery Datasets.

Area of Science:

• Computational chemistry
• cheminformatics
• Drug discovery modeling

Background:

• Hundreds of absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) and bioactivity models exist but lack public accessibility, hindering drug discovery.
• Sharing computational models remains a significant challenge in the field.

Purpose of the Study:

• To develop and release an open-source Bayesian model-building software module to improve accessibility and collaboration in drug discovery.
• To demonstrate the module's capability in generating robust predictive models for various properties.

Main Methods:

• Created a reference implementation of a Bayesian model-building module released as open-source software.
• Integrated the module into the Chemistry Development Kit (CDK), CDD Vault, and mobile applications.
• Built Bayesian models for ADME/Tox, bioactivity, and physicochemical properties using FCFP6 descriptors.

Main Results:

• The developed Bayesian models achieved cross-validation receiver operator curve values comparable to previously published models.
• The implementation enables rapid production of robust machine learning models using public or private datasets.
• Demonstrated successful model generation in proprietary software (CDD Vault) and export for open-source use (CDK).

Conclusions:

• The open-source Bayesian modeling tool enhances the accessibility and portability of predictive models in drug discovery.
• This approach facilitates biocomputation across distributed datasets, accelerating the identification of potential drug candidates.
• The integration of proprietary and open-source software promotes collaborative research and development.