OCEAN: Optimized Cross rEActivity estimatioN.

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Accurate molecular target prediction is crucial for drug discovery, aiding in off-target identification and phenotypic screen deconvolution.
• Existing methods often require extensive proprietary data, limiting accessibility.

Purpose of the Study:

• To introduce OCEAN, an open-source tool for predicting drug-target interactions using exclusively public ChEMBL database information.
• To evaluate OCEAN's performance on recently marketed drugs and novel compounds from ChEMBL releases.

Main Methods:

• OCEAN employs a heuristic approach validated against a dataset of nearly 1000 drug-target relationships.
• Prospective performance was assessed using newly released ChEMBL data (ChEMBL20 and ChEMBL21).
• The tool's capability to identify polypharmacological compounds was also evaluated.

Main Results:

• OCEAN achieved 77% success in predicting targets within the top 10 ranks for recently marketed drugs.
• For new ChEMBL20 and ChEMBL21 compounds, success rates were 62% and 51%, respectively.
• The tool correctly predicted over 64% of polypharmacological compounds within the top 10 ranks.

Conclusions:

• OCEAN provides a reliable and accessible method for drug target prediction leveraging public data.
• The tool demonstrates strong performance, particularly for known drugs, and shows promise for identifying polypharmacology.
• OCEAN's open-source availability facilitates its integration into drug discovery workflows.