We apply the concept of
proposed in ? to similarity search in protein sequences. The main question studied is the design of efficient
to construct seeds with optimal sensitivity/selectivity trade-offs. We propose several different design methods and use them to construct several alphabets. We then perform an analysis of seeds built over those alphabet and compare them with the standard
seeding method [2,3], as well as with the family of vector seeds proposed in . While the formalism of subset seed is less expressive (but less costly to implement) than the accumulative principle used in
and vector seeds, our seeds show a similar or even better performance than
on Bernoulli models of proteins compatible with the common BLOSUM62 matrix.