Abstract.
Three hundred and nine carbon-carbon, carbon-nitrogen, and carbon-oxygen p-bond lengths in high precision crystal structures of 31
purine and pyrimidine nucleobases were related to the Pauling p-bond order, its analogues corrected to crystal packing effects, the numbers
of non-hydrogen atoms around the bond, and the sum of atomic numbers of the bond atoms. Principal Component Analysis (PCA) and
Hierachical Cluster Analysis (HCA) demonstrated that the bond lengths in the nucleobases are three-dimensional phenomenon,
characterized by nine distinct classes of bonds. Bond lengths predicted by Linear Regression models, Pauling Harmonic Potential Curves,
Multiple Linear Regression, Principal Component, and Partial Least Squares Regression were compared to those calculated by molecular
mechanics, semiempirical, and ab initio methods using PCA-HCA procedure on the calculated bond lengths, statistical parameters, and
structural aromaticity indices. Incorporation of crystal packing effects into bond orders makes multivariate models to be competitive to
semiempirical results, while further improvement of quantum chemical calculations can be achieved by geometry optimization of molecular
clusters.

Keywords.
 

Keywords Plus.