Predicting the compositional evolution of the atomic-scale structure and properties of oxide glasses is important for designing new materials for advanced applications. A statistical mechanics-based approach has recently been applied to predict the composition-structure evolution in binary phosphate glasses, while topological constraint theory (TCT) has been applied in the last decade to predict the structure-property evolution in various oxide and nonoxide glass systems. In this work, we couple these two approaches to enable quantitative predictions of the compositional dependence of glass transition temperature and the population of superstructural units. The object of the study is the lithium borate glass system because they feature interesting structural characteristics (e.g., boron anomaly), and ample structure and property data are available. In these glasses, the average coordination number of boron first increases when lithium modifiers are added and then later decreases accompanied by network depolymerization. First, on the basis of B-10 nuclear magnetic resonance spectroscopy data from literature, we present a statistical description of the structural evolution in lithium borate glasses by accounting for the relative enthalpic and entropic contributions to the bonding preferences. We show that the entire glass structure evolution (both short- and intermediaterange) can be predicted based on experimental structural information for only a few glass compositions. We then show that the developed structural model can be combined with a previously established TCT model to predict the compositional evolution of the glass transition temperature. This work thus opens a new avenue for the computational design of glasses with tailored properties.