In this work, the quantitative structure–properties relationship (QSPR) was applied to modeling and predicting the 29Si-NMR chemical shifts of a series of silicate species (on Q2 sites). The descriptors that were selected by stepwise multiple linear regression technique were square of alpha polarizability, Moran autocorrelation–lag3=unweighted by atomic Sanderson electronegativities, square of asphericity, and topological path= walk 2-Randic shape index. These descriptors could encode electronic, geometric, and topological characteristics that affect the chemical shifts of the molecules of interest. The results obtained using the multiple linear regression (MLR) model were comparable with the experimental values. The cross-validation test was also performed to evaluate the prediction power of the MLR model obtained. The q2 and PRESS of this model are 0.976 and 0.44761, respectively, revealing the credibility of the model
