From activity cliffs to activity ridges: informative data structures for SAR analysis.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Extracting Structure-Activity Relationship (SAR) information from diverse compound datasets is challenging.
• Traditional SAR analysis relies on pairwise comparisons to identify "activity cliffs" (structurally similar compounds with large potency differences).
• Existing methods primarily evaluate activity cliffs at a compound-pair level, limiting comprehensive SAR insights.

Purpose of the Study:

• To extend the concept of activity cliffs by introducing "activity ridges."
• To develop a computational method for identifying and characterizing activity ridges within compound datasets.
• To provide a more systematic approach to SAR analysis and identify high-priority compound subsets for further investigation.

Main Methods:

• Defined "activity ridges" as overlapping combinatorial activity cliffs, forming ridge-like structures in activity landscapes.
• Conducted a systematic analysis across 242 compound datasets to identify activity ridges.
• Employed an information-theoretic approach to characterize the structural composition of identified activity ridges.

Main Results:

• Identified well-defined activity ridges in 71 out of 242 analyzed compound datasets.
• Demonstrated that activity ridges frequently occur in sets of active compounds.
• Showcased that different categories of activity ridges can be distinguished based on their structural content.

Conclusions:

• Activity ridges offer a rich source of SAR information, extending beyond pairwise comparisons.
• The computational identification of activity ridges provides a valuable tool for prioritizing compounds in SAR studies.
• This approach facilitates a deeper understanding of SAR determinants and accelerates drug discovery efforts.