How Reliable Are Ligand-Centric Methods for Target Fishing?

Area of Science:

• Computational chemistry
• Pharmacology
• Drug discovery

Background:

• Computational Target Fishing (TF) methods, also known as Target Prediction or Polypharmacology Prediction, are crucial for identifying new drug targets for small molecules.
• Existing TF methods often suffer from limited validation, user interpretability issues, and restricted target scope, necessitating improved evaluation strategies.
• Target-centric TF methods have inherent limitations in the number of predictable targets compared to ligand-centric approaches.

Purpose of the Study:

• To propose and introduce a novel benchmark for validating Target Fishing (TF) methods.
• To analyze the variability of predictive performance in TF methods based on query molecules.
• To assess the current state of polypharmacology and the predictability of approved drug targets.

Main Methods:

• Development of a new benchmark specifically designed to evaluate TF method performance across different query molecules.
• Analysis of approved drugs to estimate the number of predicted targets requiring testing to identify true targets with submicromolar potency.
• Comparison of target prediction difficulty between approved drugs and randomly selected molecules using a control group.

Main Results:

• On average, testing five predicted targets is estimated to be sufficient to identify two true targets with submicromolar potency for approved drugs, though performance varies significantly.
• Approved drugs are found to have an average of eight known targets, supporting the prevalence of polypharmacology.
• Targets of approved drugs are generally more challenging to predict compared to targets of randomly selected molecules.

Conclusions:

• The proposed benchmark offers a more robust and interpretable validation framework for Target Fishing (TF) methods.
• The findings underscore the common occurrence of polypharmacology and provide insights into the efficiency of TF methods in drug discovery.
• The study highlights the need for improved TF methods and validation strategies to better predict drug-target interactions and facilitate drug repositioning.