Quantitative Structure Property Relationship (QSPR) analysis based on molecular dynamics (MD) simulations is a promising approach for establishing the composition-property relationships of glass and other materials with complex structures. A series of twenty borosilicate and boroaluminosilicate glasses have been modeled using MD simulations with recently developed effective potentials. Short- and medium-range structures of these glasses were analyzed and, based on the structural information, QSPR analysis of the initial dissolution rate (r0 ) was made and compared with measured r 0 at 90°C and pH 9 using various structural descriptors such as percentage of bridging oxygen species, network connectivity and average ring size. The structural descriptors, Fnet , containing energetic information such as single bond strength and other structural information were also used. It was found that overall network connectivity, average ring size and F net give reasonable predictions of the r 0 of studied glasses, given the conditions that the glasses are homogeneous and dissolve congruently. Modifying glass compositions to account preferential release of modifiers gives a better prediction for incongruently dissolving glasses. The findings were discussed with results of recent studies on evaluating the compositional dependence of glass dissolution behavior using the topological-constraints-based models.