Methodological aspects of conformational search and entropy calculation in gas-phase vitamers

The systematic, Uniconf-based search, as well as the CREST metadynamics simulations, combined with a number of contemporary semiempirical methods (GFNn-xTB (n = 1, 2) and QM3) have been employed for the conformational sampling and entropy evaluation of vitamins A, B1, B2, B3, B5, B6, B7, B9, and C. Using the tightest settings, both the Uniconf and CREST strategies are capable of finding the most ΔE stable conformer within the same semiempirical potential. The density functional theory (DFT) refinement of the most stable structures indicates the following: (a) there is no single semiempirical method that permanently reproduces the DFT conformational preferences, as when one fails, another works; (b) on average, the most ΔG stable conformer obtained with a given semiempirical method is preferable over its ΔE counterpart as an approximation to the most ΔG stable DFT conformer. The conformational entropy for a particular vitamin depends on the conformational sampling technique as well as on the semiempirical method used. In the case of the CREST entropy mode, a few conformational entropies have been found to be atypically large. The origin of these anomalies is rooted in the difference between the Boltzmann-averaged molecular entropy and its counterpart obtained for the most stable conformer . All the utilized approaches to the calculation of conformational entropy should only be considered estimates. These are still of practical use in the absence of high-quality data, which are expected to be obtained from the DFT treatment of the complete conformational space generated either with CREST or Uniconf.