The dynamics of Chlorophyll a (CLA) aggregation in the plant thylakoid membranes have been investigated using coarse‐grained molecular dynamics simulations and machine learning approaches at K. The CLA molecules dynamically form and break CLA dimers and higher‐order aggregates. The contact lifetime and waiting time distributions of CLA dimers exhibit the coexistence of the fast and slow time scales. The survival probability of CLA dimers follows a non‐exponential decay with multiple residence time scales and a time‐dependent rate, unlike conventional rate theory, manifesting the emergence of disorder. Markov state modeling using an autoencoder‐derived latent representation of pairwise distances among CLAs results in three discrete interconvertible metastable states with different conformational distributions during the aggregation process. The mean first‐passage times between Markov states of different conformations are slower than the longest residence time of the dimers. This demonstrates that these transitions represent the slowest kinetic processes during aggregation. Our results provide the underlying mechanism of membrane‐bound chlorophyll aggregation, which will have useful applications in photodynamic therapy and artificial photosynthetic materials in the future.
Saini et al. (Sat,) studied this question.