Using biased tetrapeptide libraries made up of proteinogenic amino acids of the general formula Cys-O2-X3-X4, we searched for new substrates of partly purified rat brain S-farnesyl transferase (FTase). To achieve this task, an assay was developed in which the consumption of the co-substrate (farnesyl pyrophosphate) was measured. After three steps of deconvolution including each synthesis and enzymatic assay, the most efficient substrates found under these particular conditions were Cys-Lys-Gln-Gln (peptide I) and Cys-Lys-Gln-Met (peptide II). As a control, we used another tetrapeptide library (Cys-Val-O3-X4) in which the valine position was arbitrarily fixed, corresponding to Cys-Val-Ile-Met in the CAAX box of K-RasB, although this sublibrary was only marginally active compared with Cys-Lys-X3-X4 in the first round of deconvolution. The best substrate sublibrary was Cys-Val-Thr-X4, threonine being more favourable than the aliphatic amino acids (Val, Ile, Leu, Ala) in this position. Deconvolution finally led to Cys-Val-Thr-Gln, -Met, -Thr and -Ser as the most efficient substrates of FTase. Those tetrapeptides were not substrates of a partly purified geranylgeranyl transferase 1 (GGTase1). We also investigated the influence of the -1 position (at the N-terminus of cysteine) on the specificity of the enzyme, by using a series of pentapeptides constructed on the basis of the best tetrapeptide core (peptide 1). Among this family of analogues, only His-Cys-Lys-Gln-Gln did not behave as a substrate, whereas all the other pentapeptides were measurable substrates, with Gly-, Asn- and Thr-Cys-Lys-Gln-Gln displaying kinetic constants similar to that of Cys-Lys-Gln-Gln. The present work provides strong evidence that the best tetrapeptide substrates of FTase do not necessarily belong to the classical CAAX box, in which A's are lipophilic residues, but rather contain hydrophilic amino acids in the middle of their sequences. Among them, peptides I and II are potent FTase in vitro substrates that are not recognised by GGTase1 and might be new starting points for the design of FTase inhibitors.