Bioalerts: a python library for the derivation of structural alerts from bioactivity and toxicity data sets.

Area of Science:

• Computational chemistry
• Drug discovery
• Toxicology

Background:

• Early assessment of compound toxicity is vital in drug discovery to prevent clinical trial failures.
• Identifying structural alerts, chemical fragments linked to toxicity, is a key strategy.
• Automated methods for deriving structural alerts from toxicity data have been developed over decades.

Purpose of the Study:

• To present the Python library bioalerts for automated structural alert derivation.
• To enable analysis of both categorical and continuous bioactivity data.
• To provide tools for chemical space deconvolution and visualization.

Main Methods:

• Utilizes the RDKit implementation of the circular Morgan fingerprint algorithm.
• Computes chemical substructures based on radial atom neighborhoods and increasing bond radii.
• Applies algorithms for automatic derivation of structural alerts from diverse datasets.

Main Results:

• The bioalerts library facilitates the derivation of structural alerts from dichotomous and continuous bioactivity data.
• It enables calculation of unhashed Morgan fingerprints for predictive modeling.
• Provides functionalities for straightforward visualization of derived structural alerts.

Conclusions:

• bioalerts offers a comprehensive solution for identifying potential toxic liabilities in drug candidates.
• The library supports advanced cheminformatics analyses, including predictive modeling and chemical space exploration.
• Facilitates early-stage toxicity assessment, streamlining the drug discovery pipeline.