### abstract

- We formalize a new problem variant in gene-block discovery, denoted Reference-Anchored Gene Blocks (RAGB), given a query sequence Q of length n, representing the gene array of a DNA element, a window size bound d on the length of a substring of interest in Q, and a set of target gene sequences [Formula: see text]. Our objective is to identify gene blocks in [Formula: see text] that are centered in a subset q of co-localized genes from Q, and contain genomes from [Formula: see text] in which the corresponding orthologs of the genes from q are also co-localized. We cast RAGB as a variant of a (colored) biclique problem in bipartite graphs, and analyze its parameterized complexity, as well as the parameterized complexity of other related problems. We give an [Formula: see text] time algorithm for the uncolored variant of our biclique problem, where m is the number of areas of interest that are parsed from the target sequences, and n and d are as defined earlier. Our algorithm can be adapted to compute all maximal bicliques in the graph within the same time complexity, and to handle edge weights with a slight [Formula: see text] increase to its time complexity. For the colored version of the problem, our algorithm has a time complexity of [Formula: see text]. We implement the algorithm and exemplify its application to the data mining of proteobacterial gene blocks that are centered in predicted proteobacterial genomic islands, leading to the identification of putatively mobilized clusters of virulence, pathogenicity, and resistance genes.