Fig. 3

The simplified binning scheme and search strategy implemented in CouchDB. For the genomic feature A, the smallest containing bin is the one reached by navigating the tree in the following way: 0,0,1,0,0 (in red). For Feature B and Feature U the smallest bins are (0,1) and (0,0,1,1,1) respectively. Given the interval query Q, its smallest containing bin is the one coded by (0,0,1). When searching for genomic features within the corresponding overlapping bins, both for the lower and upper part of the tree, genomic feature U would also be reported: in fact, despite overlapping with one of the searched bins, it does not overlap with Q. Therefore, two more queries (views) are performed in order to remove the non-overlapping elements: the first adds the start position of the genomic feature to the view keys (patient id, chromosome, bin, start) while the second one adds the stop position. In this example, genomic feature U would be removed from the query result set because its start position is greater than the end one of Q