Data-informed reparameterization of modified RNA and the effect of explicit water models: Application to pseudouridine and derivatives

Pseudouridine is the most abundant post-transcriptional modification in RNA. We have previously shown that the FF99-derived parameters for pseudouridine and some of its naturally occurring derivatives in the AMBER distribution either alone or in combination with the revised 𝛄 torsion parameters (parmbsc0) failed to reproduce their conformational characteristics observed experimentally (Deb I, et al. J. Chem. Inf. Model. 2014, 54 (4):1129–1142; Deb I, et al. J. Comput. Chem., 2016, 37:1576−1588; Dutta N, et al. J. Chem. Inf. Model. 2020, 60 (10):4995–5002). However, the application of the recommended bsc0 correction did lead to an improvement in the description not only of the distribution in the 𝛄 torsional space but also of the sugar pucker distributions. In an earlier study, we examined the transferability of the revised glycosidic torsion parameters (𝛘IDRP) for Ψ to its derivatives. We noticed that although these parameters in combination with the AMBER FF99-derived parameters and the revised 𝛄 torsional parameters resulted in conformational properties of these residues that were in better agreement with experimental observations, the sugar pucker distributions were still not reproduced accurately. Here we report a new set of partial atomic charges for pseudouridine, 1-methylpseudouridine, 3-methylpseudouridine and 2′-O-methylpseudouridine and a new set of glycosidic torsional parameters (𝛘ND) based on chosen glycosidic torsional profiles that most closely corresponded to the NMR data for conformational propensities and studied their effect on the conformational distributions using REMD simulations at the individual nucleoside level. We have also studied the effect of the choice of water model on the conformational characteristics of these modified nucleosides. Our observations suggest that the current revised set of parameters and partial atomic charges describe the sugar pucker distributions for these residues more accurately and that the choice of a suitable water model is important for the accurate description of their conformational properties.