2D QSAR studies were carried out for a series of 55 ligands for the Thyroid receptors, TRalpha and TRbeta. Significant cross-validated correlation coefficients (q(2)=0.781 (TRalpha) and 0.693 (TRbeta)) were obtained. The models' predictive abilities were proved more valuable than the classical 2D-QSAR, and were further investigated by means of an external test set of 13 compounds. The predicted values are in good agreement with experimental values, suggesting that the models could be useful in the design of novel, more potent TR ligands. Contribution map analysis identified a number of positions that are promising for the development of receptor isoform specific ligands.