Mimicry epitopes that are recognized by T-cells can be identified through screening of synthetic peptide libraries. We have shown that these mimicry epitopes share sequence similarity with the corresponding natural epitopes and that mimicry sequences can be used for the definition of protein derived T-cell epitopes from databases. This can be done by either homology searching or pattern searching. Here we discuss the advantages and disadvantages of homology searching as an alternative for the generally applicable recognition pattern approach. We show that only for part of the library derived mimicry epitopes, the degree of similarity to the natural epitope may be high enough for successful homology searching in small databases.