Challenges with multi-objective QSAR in drug discovery

ABSTRACT

Introduction: The complexity in the drug discovery pipeline, in combination with the exponential growth of experimental and computational data, the technological achievements, and the access to large data sets, has led to a continuous evolution and transformation of quantitative structure–activity relationships (QSAR) to compete with the challenges of multi-objective drug discovery.

Areas covered: After a short overview of the multiple objectives involved in drug discovery, this review focuses on definition of the drug-like space and the construction of local and/or global models, platforms and workflows for step-by-step single-objective optimization (SOO) of the different and often conflicting processes. Multi-targeted drug design is a particular case of multi-objective QSAR integrated into the new era of polypharmacology. Multi-objective optimization (MOO), based on desirability functions or Pareto surfaces and its application in QSAR, as an alternative optimization philosophy, is also discussed.

Expert opinion: Access to large databases as well as to software services by means of cloud technology facilitates research for more efficient and safer drugs. QSAR models implemented in web platforms and workflows provide sequential SOO for multiple biological and toxicity end points, while MOO, still restricted to a limited number of objectives, is helpful for multi-target or selectivity design, as well as for model prioritization.

KEYWORDS:

• QSAR
• multi-objective drug design
• multi-target drug design
• drug-like space
• single objective optimization
• multi-objective optimization

Article highlights

• Drug discovery is a complex process involving multiple and often conflicting objectives as well as multiple potential causes of candidate attrition.

• Quantitative structure–activity relationships (QSAR) has adapted to the multitask concept moving from the perception of single-objective to the multi-objective drug discovery.

• Drug-likeness defines boundaries on the chemical space and functions as filter to guarantee a physicochemical profile enabling further development.

• Predictions of potency and the different biological/pharmacokinetic/toxicity end points rely on the construction of local or global QSAR models, web platforms and workflows.

• Step-by-step single-objective optimization considers each property or each model separately.

• Multi-objective optimization is based on overall property scores, desirability functions, and the Pareto concept, permitting to conceive an ‘overall QSAR’ as a function of the separate models.

• Further challenges for QSAR are how to address the multiple pathways of complex diseases, the targeted therapies for individual population groups, and the treatment of big data.

This box summarizes key points contained in the article.

Declaration of Interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This manuscript was not funded.