Mining the literature to map which substituents medicinal chemists swap for one another
- Substituent pairs are treated as bioisosteres when they appear in the same position of related compounds that act on the same target, and these pairs are linked into a network of interchangeable groups.
- Data come from compounds reported in the medicinal chemistry literature; the approach relies on co-occurrence in matching molecular contexts instead of similarity or property descriptors.
- It reflects substitutions chemists actually made and surfaces lesser-known replacements; the network is offered as a free web tool, with no experimental testing described.
💡 Points Based on the abstract
Co-occurrence mining encodes what chemists chose to make, not what worked: papers report successes, so the network inherits publication bias and says nothing about potency cliffs or ADME changes after a swap. Useful as an idea generator beyond a Topliss-style tree, but I'd want pairs ranked by measured activity deltas, not frequency, before trusting it on my series.
📄 Paper · Journal of Chemical Information and Modeling (ACS) ↗ A Cheminformatics Analysis of Substituents in Bioactive Molecules and Their Bioisosteric Replacement Network