Data-driven assessment of bioisosteric replacements and their influence on off-target activity profiles

Abstract

Bioisosterism, a fundamental concept in medicinal chemistry, involves the substitution of chemical groups with structural analogs that preserve similar physicochemical properties while potentially modulating potency or toxicity. To systematically investigate shifts in pChEMBL values upon such substitutions, we developed a KNIME workflow that extracts and analyzes compound pairs featuring literature-curated common bioisosteric exchanges. The workflow retrieves pChEMBL values across 88 off-targets from ChEMBL and supports decision-making through pair-level quality metrics such as the document consistency ratio and assay context consistency ratio, which assess the consistency of the source data. Our analysis revealed that ester-to-secondary-amide replacements at the muscarinic acetylcholine receptor M2 (CHMR2) result in a significant mean decrease in pChEMBL of 1.26 across 14 compound pairs (p < 0.01). In contrast, phenyl-to-furanyl substitutions at the adenosine A2A receptor (ADORA2A) led to a mean increase in pChEMBL of 0.58 across 88 compound pairs (p < 0.01). Furthermore, a second KNIME workflow was developed to assess selectivity profiles by analyzing pChEMBL shifts at secondary targets. Among 66 compound pairs active at both ADORA2A and ADORA1, the mean change at ADORA1 was only +0.14 ± 0.52, indicating a selective potency increase at ADORA2A. This exemplifies a potential case of increased potency at an off-target associated with adverse effects, while maintaining activity at a pharmacologically desirable target. Conversely, furanyl-to-phenyl replacements may selectively reduce undesired potency at ADORA2A while preserving potency at ADORA1. This framework enables systematic, data-driven evaluation of potency shifts induced by bioisosteric replacements, aiding in the identification of substitutions associated with off-target potency increases or decreases during lead optimization. The workflow offers a semi-automated, reproducible approach that integrates bioisostere generation, activity mapping, and statistical assessment in a single platform, making it readily adaptable to other compound series and target panels. In addition, it evaluates whether activity at other known targets remains unchanged, thereby providing an assessment of selectivity of the replacements. The workflow can be applied to prioritize replacement strategies that reduce off-target risks, evaluate selectivity profiles, and generate curated potency shift data to support predictive modeling efforts.