Abstract:

Based on a novel molecular distance-edge vector (VMDE,ν), expressed in the form of potential energy and developed in our laboratories, further syst ematic studies were made on the regularity of chemical shift sum (CSS) for c arbo n-13 nuclear magnetic resonance ¹³C NMR).  With aid of several chemom etrical techniques including multiple linear regression (MLR), stepwise multiple regressi on (SMR) and principal component regression (PCR), various molecular modelling e quations were established to correlate chemical shift sum (CSS) of carbon-13  nucl ear magnetic resonance ¹³C NMR) excellently to the molecular distance-edge vec tor (VMDE) together with the path count of three bonding segment (C-C-C) as foll ows, whose very good results were obtained for the modelled estimation and accur ate prediction with cross validation (CV) of leave-one-out (LOO) procedure: CSS=bν+cp₃=∑^m_j=0b_jν_j+b₁₁p₃=b₀ν+b₁ν₁+b₂ν₂+b₃ν₃+b₄ν₄+b₅ν₅+b₆ν₆+b₇ν₇+b₈ν₈+b₉ν₉+b₁₀ν₁₀+b₁₁p₃=-13.576+22.179ν₁+28.407ν₂+25 .950ν₃+26.690ν₄+14.498ν₅+5.726ν₆-5.379ν₇-3.214ν₈-15.021ν₉ -25.710ν₁₀+12.278p₃  n=63, R=0.997, EV=99.68%, RMS=3.7348, SD=4.1 18, F= 773.116, U=144228.844, Q=864.938; CV: R²_CV=0.980, EV=98.83%, RMS=7.126 1, SD_CV=7.634, F_CV=221.720, U_CV=142121.891, Q_CV=2971 .896. Some reliable correlation models have been developed by using the adjoin structural descriptors through combination of the MDE vector (ν vector) and molecu lar path counts of length three (p₃).

Key words: Molecular modelling and simulation, Molecular distance-edge vector ν, Chemical structural parameterization, Chemometrics and chemoinformatics, Multip le linear regression (MLR), Stepwise multiple regression (SMR), Principal comp onent regression (PCR), Carbon-13 nuclear magnetic resonance spectroscopy ¹³C NMR), Cross validation, Chemical shift sum of alkanes, Quantitative struc ture-spectrum relationships (QSSR)