Mining Chemical Activity Status from High-Throughput Screening Assays.

Area of Science:

• Computational chemistry
• Cheminformatics
• Bioinformatics

Background:

• High-throughput screening (HTS) generates vast data on chemical compound activity against molecular targets.
• Predicting compound activity (active/inactive) in HTS assays is challenging due to large datasets and imbalanced class distributions (few active compounds).

Purpose of the Study:

• To develop and validate a novel computational method, DRAMOTE, for accurate prediction of chemical compound activity in HTS assays.
• To demonstrate DRAMOTE's superior performance compared to existing state-of-the-art methods.

Main Methods:

• DRAMOTE was developed by modifying a minority oversampling technique to handle imbalanced datasets common in HTS.
• Performance was evaluated through comprehensive comparative analysis against other methods using data from 11 PubChem assays (approx. 500,000 interactions).
• The method was applied to predict FDA-approved drugs interacting with the thyroid stimulating hormone receptor (TSHR).

Main Results:

• DRAMOTE achieved considerably better results than current state-of-the-art solutions for a class of HTP assays.
• Comprehensive analysis across 1,350 experiments involving ~500,000 chemical-protein interactions demonstrated DRAMOTE's superior performance.
• Application to TSHR prediction yielded robust models, partially supported by 3D docking and literature data.

Conclusions:

• DRAMOTE is a highly effective method for predicting chemical compound activity in HTS assays, outperforming existing approaches.
• The method can be utilized for developing robust virtual screening models in drug discovery.
• Datasets and implementation are available as a MATLAB toolbox.