Badapple 2.0: An Empirical Predictor of Compound Promiscuity, Updated, Modernized, and Enhanced for Explainability

Area of Science:

• Cheminformatics
• Biomedical Data Science
• Drug Discovery

Background:

• Bioassay data analysis is crucial for modern drug discovery and chemical biology.
• Identifying false trails from promiscuous compounds is a significant challenge.
• The original Badapple tool (BioAssay-Data Associative Promiscuity Pattern Learning Engine) was released in 2012 to address this issue.

Purpose of the Study:

• To introduce Badapple 2.0, a major update to the Badapple software.
• To enhance functionality, scalability, and data handling for bioassay analysis.
• To improve the identification of promiscuous compounds and reduce false trails in drug discovery.

Main Methods:

• Complete code rewrite incorporating software engineering, cheminformatics, and biomedical data science.
• Updated and expanded assay data sets.
• Enhanced metadata for improved explainability and richer bioactivity analyses.

Main Results:

• Badapple 2.0 offers enhanced functionality, scalability, and metadata.
• The updated tool supports improved explainability and richer bioactivity analyses.
• The system has broad applicability for early-stage drug discovery campaigns.

Conclusions:

• Badapple 2.0 represents a significant advancement in analyzing bioassay data.
• The tool is vital for identifying promiscuous compounds and avoiding false trails.
• This updated software improves the efficiency of drug discovery research, particularly in AI/ML-empowered programs.