基于NMR理论计算的参数优化及其在天然产物中的应用OA

This study investigates the predictive performance of various functional/basis set combinations in nuclear magnetic resonance(NMR)quantum-chemical calculations.The training set CSRT80,the validation set CSRP25,three natural products with crystallographic diffraction data and eight compounds included in the custom DP4+model are selected.The effects of different functional/basis set combinations for geometry optimization and NMR calculations on the accuracy and computational cost of 1H-NMR and 13C-NMR chemical shift predictions are evaluated.The results indicate that,no single functional/basis set combination simultaneously achieves high prediction accuracy and low computational cost.Nevertheless,the revTPSS/pcSseg-1 combination demonstrates the best balance between accuracy for 1H-NMR and 13C-NMR chemical shift predictions,yielding δH-MAD of 0.289 9×10-6 and δC-MAD of 1.867 9×10-6,with computational cost significantly lower than the sub-optimal combinations(TPSS/pcSseg-1,TPSSh/cc-pVTZ,and revTPSS/cc-pVTZ).Geometry optimization at the r2SCAN-3c level,combined with NMR chemical shift calculations using the revTPSS/pcSseg-1 combination and scaled correction of NMR data,achieves a good balance between prediction accuracy and computational cost.