From handbooks to high-throughput: rule-based prediction of electronic absorption maxima from SMILES with ChromoPredict

Abstract

Accurate prediction of electronic absorption spectra is essential for the rational design of photofunctional molecules. While ab initio quantum chemical methods provide reliable results, their high computational cost often precludes their application in high-throughput or resource-constrained screening workflows. Data-driven alternatives can offer improved efficiency but typically require large, high-quality datasets and may lack interpretability. In this work, we present a low-cost, interpretable approach for predicting absorption maxima (λ_max) based on digitized and extended empirical rules originally proposed by R. B. Woodward, M. Fieser, L. Fieser and H. Kuhn. These rule sets estimate ππ* transition energies through additive contributions from base chromophores and position dependent contributions of certain structural features and substituents. Our implementation enables direct prediction of λ_max from SMILES input for three representative compound classes: (i) α, β-unsaturated carbonyl compounds, for which we introduce a refined rule set, (ii) dienes and polyenes, and (iii) 3,4,6-substituted coumarin derivatives. For the latter, we define an entirely new set of empirical rules based on literature data. The resulting workflow offers a computationally efficient and chemically interpretable alternative for early-stage molecular screening and design, bridging historical empirical knowledge with modern cheminformatics.