A homology retrieval can be performed with higher accuracy than conventional technologies when comparing a query sequence with a target sequence, and retrieving a similar location in the target sequence. The sequence information of a query sequence and a genomic-scale target sequence is acquired, the acquired information is compressingly converted into a compressed query sequence and a compressed target sequence in each of which a homopolymer region including two or more consecutive identical bases is replaced with a single base of the bases, the two sequences are compared, and a refining search is performed for a compressed target partial sequence that matches the compressed query sequence in the compressed target sequence. For the refined compressed candidate sequence and the query sequence, based on the information on the number of consecutive identical bases in the each of the sequences before compression, the number of consecutive bases is compared between the two compressed sequences for each corresponding base, and the degree of similarity indicating homology of the candidate sequence with the query sequence is computed from a degree of match or a degree of mismatch in the number of consecutive bases. By ranking and selecting an arbitrary number of candidate sequences having relatively high homology with the query sequence from this degree of similarity, it is possible to avoid the influence of the number of consecutive identical bases in a homopolymer region, thereby performing a homology retrieval accurately.