Exploring the biologically relevant chemical space for drug discovery

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Pharmacology

Background:

• Physicochemical properties of launched drugs have evolved over decades.
• Drug discovery trends show decreased biological relevance (BR) and similarity to natural products (NPs).
• Current high-investment, low-productivity drug discovery necessitates improved screening and exploration of biologically relevant chemical spaces.

Purpose of the Study:

• To address the declining biological relevance in drug discovery.
• To introduce a novel descriptor, the biologically relevant spectrum (BRS), for quantitative structure-activity relationships (QSAR) and library preparation.
• To develop predictive models for biological activity and pharmacokinetic properties.

Main Methods:

• Analysis of physicochemical properties of launched drugs.
• Development of the biologically relevant spectrum (BRS) descriptor.
• Creation of prediction models using BRS and support vector machine (SVM) for 43 biological activity categories.
• Validation of models on drugs launched between 2006-2012.

Main Results:

• Launched drugs show decreasing biological relevance and similarity to natural products.
• BRS descriptor proposed for QSAR and screening library design.
• SVM models achieved high prediction accuracies (around 95%) and MCC values (>0.8) for 43 biological activities.
• Successful prediction of 37 out of 48 drug-activity associations for recent drugs.

Conclusions:

• Drug discovery should shift focus back towards natural products to enhance biological relevance.
• BRS is a valuable tool for improving screening library design and drug discovery efficiency.
• Developed predictive models and the BioRel web server offer practical applications for drug discovery research.