The amino acid sequences flanking 352 dibasic moieties contained in 83 prohormones and pro-proteins listed in a database were examined. Frequency calculations on the occurrence of given residues at positions P6 to P'4 allowed us to delineate a number of features which might be in part responsible for the in vivo discrimination between cleaved and uncleaved dibasic sites. These include the following: amino acids at these positions were characterized by a large variability in composition and properties; no major contribution of a given precursor subsite to endoprotease specificity was observed; some amino acid residues appeared to occupy preferentially certain precursor subsites (for instance, Met in P6 and P3, Asp and Ala in P'1, Pro in P6, Gly in P3 and P'2 etc.) whereas some others appeared to be excluded. Most amino acid residues occupying the P'1 position in these precursor cleavage sites were tolerated. But the beta-carbon branched side chain residues (Thr, Val, Leu, Ile) and Pro, Cys, Met and Trp were either totally excluded or poorly represented, suggesting that they might be unfavourable to cleavage. The biological relevance of these observations to the efficacy of dibasic cleavage by model propeptide convertases was in vitro tested using both pro-ocytocin convertase and Kex2 protease action on a series of pro-ocytocin related synthetic substrates reproducing the Pro7-->Leu15 sequence of the precursor in which the Ala13 residue (P'1 in the LysArg-Ala motif) was replaced by various amino acid residues. A good correlation was obtained on this model system indicating that P'1 residue of precursor dibasic processing sites is an important feature and may play the role of anchoring motif to S'1 convertase subsite. We tentatively propose that the present database, and the corresponding model, may be used for further investigation of dibasic endoproteolytic processing of propeptides and pro-proteins.