High-throughput DFT screening of glutathione-triggered drug release from Au13, Ag13, and Pt13 nanoclusters: metal-level selectivity and the limits of geometry-level screening.

Abstract

Glutathione (GSH)-responsive noble-metal nanoclusters exploit the ∼1000-fold intracellular GSH gradient between tumour and healthy tissue for selective drug release, but rational design of drug-metal pairings has lacked systematic binding data across chemical space. We screened 81 anticancer and drug-like molecules on icosahedral Au13, Ag13, and Pt13 nanoclusters (238 complexes) using dispersion-corrected density functional theory, mapping binding energies onto Boltzmann-weighted tumour selectivity indices. Charge and bond-order analysis reveals a mechanistic dichotomy: Au and Ag coordinate drugs through heteroatom σ-donation (N, O, S), whereas Pt forms η2-π complexes through carbon. Pt binds drugs at -61.4 kcal mol-1 on average, far beyond the strongest heteroatom donation in the dataset (≈-44 kcal mol-1), so no thiol competitor can displace π-bound cargo; Pt is therefore unsuitable as a GSH-triggered carrier, a conclusion robust to spin state, cluster charge, cluster size, basis-set superposition error and level of theory. Individual drug rankings are not. A 5.3 kcal mol-1 displacement-energy window defines tumour-selective release, and 32 complexes fall inside it at the screening level, but geometry relaxation, larger clusters, and corrected thiol-binding modes each dismantle these nominations; the combined uncertainties exceed the window itself, so the window is not resolvable at this level of theory. Machine learning reproduces the metal-level dichotomy but predicts within-carrier release only weakly. Carrier choice, not drug choice, is the reliable design variable, and candidate-level selectivity claims require full geometry optimisation and free-energy corrections.

Abstract as supplied by the indexing source. © the authors / publisher.