A new paper from our group, co-authored with Parisa Fereidounpour and Kim Lambertsen Larsen, has been published in the Journal of Inclusion Phenomena and Macrocyclic Chemistry.

The anilinonaphthalene sulfonates 2,6-ANS and 8,1-ANS are polarity-sensitive fluorescent probes, and their sensitivity to the local environment makes them attractive as cheap, high-throughput reporters for cyclodextrin inclusion complexes. Using them that way requires trustworthy reference binding free energies, so we measured the association of both probes to α-, β-, and γ-cyclodextrin by fluorescence titration and isothermal titration calorimetry, and compared the results against docking and MM/GBSA calculations.

The methods do not agree as well as one would like. Fluorescence points to β-CD as the strongest host for 2,6-ANS and γ-CD for 8,1-ANS, but calorimetry ranks the 8,1-ANS–γ-CD complex among the weakest in the set. The two techniques agree closely where binding is strongest — 2,6-ANS with β-CD gives ΔG_a of −18.9 kJ/mol by fluorescence and −19.4 kJ/mol by ITC — yet diverge by 6.6 kJ/mol for 2,6-ANS with α-CD. Docking reproduces the experimental ranking qualitatively, while MM/GBSA fails systematically for γ-CD, returning positive binding free energies; only 3–13 of 30 γ-CD trajectories met our convergence criteria, against 17–30 for α- and β-CD.

Each method fails for its own reason. Molecular dynamics shows that the naphthalene ring of 2,6-ANS fits poorly in the α-CD cavity and stays partially solvent-exposed, biasing the extracted association constants downward; for the bulkier 8,1-ANS, γ-CD is the only host studied that fully accommodates the naphthalene group. The weak complexes also fall outside the reliable Wiseman window for ITC, where K_a and ΔH_a become strongly correlated and hard to separate. And the surface-area-based nonpolar solvation term in MM/GBSA breaks down when water shares a confined cavity with the ligand.

The conclusion is that no single method reliably quantifies binding in this regime (K_a below roughly 5 × 10³ M⁻¹), and that progress needs cyclodextrin-tailored computational models with nonpolar solvation terms reparameterised against experimental benchmarks. The data are best read as a methodological comparison rather than a definitive reference set for ANS–cyclodextrin thermodynamics.