Proteochemometrics (PCM) is a method for deriving quantitative structure–activity relationships (QSAR). It models the bioactivity of multiple ligands against multiple target proteins. Thus, PCM is a key method used to develop multi-target drug molecules. In the present protocol, the PCM method is applied to a set of peptides binding to seven polymorphic HLA class II proteins from locus DP. The peptides bind to an open-ended binding site on DP proteins which accepts a nonameric binding core. As the peptides in the studied set are 15-mers, the initial set is processed by an iterative self-consistent algorithm to select the most probable binding nonamer for each 15-mer peptide. The final set containing the most probable binding nonamers is used to derive the PCM model. The model is validated by external set of proteins and enters the server EpiTOP which is freely accessible at http://www.ddg-pharmfac.net/EpiTOP3/.