Infinite dilution gas solubility coefficients in glassy polymers of diverse chemical structure are predicted on the basis of a simple computational model of short polymer chain conformations. In contrast to existing models that describe the infinite dilution solubility coefficient S of various gases in certain polymers or S for the particular gas in different polymers, our approach allows to deal with the broad variety of gases and glassy polymers of different structure. With the van der Waals volume, accessible surface area, and some derivative geometrical descriptors being the predicting variables, the multivariate linear regression on the testing set gives correlation 0.89 with the logarithm of the solubility coefficient for the considered classes of relatively low-permeable polymers (the mean relative error of solubility coefficient S prediction is 68%). Regression accuracy can be improved up to correlation 0.95 and the mean relative error 39% when a separate regression is constructed for each penetrant gas. Implications are discussed to the analysis of solubility selectivity. (c) 2016 Wiley Periodicals, Inc. J. Polym. Sci., Part B: Polym. Phys. 2017, 55, 228-244