The published data for six melanocortin peptides binding to wild-type and chimeric melanocortin MC(1)/MC(3) receptors were analysed using the novel proteo-chemometrics modelling approach. The chimeric receptors and the peptides were coded using binary descriptors and used to correlate with the experimental data for affinity or selectivity for peptides binding to receptors. Correlations were achieved using partial least squares projection to latent structures (PLS) and statistically valid models were obtained. The models were further improved by adding cross-terms and applying orthogonal signal correction. The models were validated using external prediction, with half of the data being excluded from the modelling. Interpretation of the results using PLS coefficient plots revealed that the binding pocket for the melanocortins is located between the first, second, third, sixth and seventh transmembrane regions of the melanocortin receptors, in good agreement with previous three-dimensional models for the interactions of melanocortins with melanocortin receptors. Further, analysis of cross-terms between peptide descriptors indicated that the proteo-chemometrics modelling is able to distinguish between differences in the conformational space of the peptides that affect binding affinity and selectivity.