Machine Learning Assisted Hit Prioritization for High Throughput Screening in Drug Discovery

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• High throughput screening (HTS) is crucial for drug discovery but often generates false positives due to assay interference.
• Efficiently prioritizing true bioactive compounds from noisy HTS data remains a significant challenge.

Purpose of the Study:

• To develop a data-driven approach for simultaneous detection of assay interferents and prioritization of true bioactive compounds.
• To provide a universally applicable method for any HTS campaign without prior knowledge of interference mechanisms.

Main Methods:

• Utilized a gradient boosting model trained on noisy HTS data.
• Introduced a novel formulation of sample influence to analyze model learning dynamics.
• Distinguished between true bioactivity and assay artifacts based on learning patterns.

Main Results:

• The method successfully identified and excluded assay interferents with diverse mechanisms.
• Demonstrated superior prioritization of biologically relevant compounds compared to existing baselines.
• Achieved high computational efficiency, processing assays in under 30 seconds on low-resource hardware.

Conclusions:

• The developed approach offers a robust and efficient solution for false positive detection in HTS.
• This tool can significantly enhance the efficiency of drug discovery pipelines by guiding pharmacological optimization post-HTS.