Rather than using a fixed set of amino acid frequencies, modify tide-index to store the amino acid frequencies of the database. At search time, combine those frequencies with a stored frequency vector derived from NRDB, using a user-specified pseudocount to weight the two.