A semi-empirical method based on the mass-per-flexible-bond (M/f) principle was used to quantitatively explain the large range of glass transition temperatures (T-g) observed in a library of 132 L-tyrosine derived homo, co- and terpolymers containing different functional groups. Polymer class specific behavior was observed in T-g vs. M/f plots, and explained in terms of different densities, steric hindrances and intermolecular interactions of chemically distinct polymers. The method was found to be useful in the prediction of polymer T-g. The predictive accuracy was found to range from 6.4 to 3.7 K, depending on polymer class. This level of accuracy compares favorably with (more complicated) methods used in the literature. The proposed method can also be used for structure prediction of polymers to match a target T-g value, by keeping the thermal behavior of a terpolymer constant while independently choosing its chemistry. Both applications of the method are likely to have broad applications in polymer and (bio)material science. (C) 2007 Elsevier Ltd. All rights reserved.