Approximate pattern matching using search schemes and in-text verification

Search schemes enable the efficient identification of all approximate occurrences of a search pattern in a text. Using a bidirectional FM-index, search schemes describe how to explore the search space in such a way that runtime is minimized. Even though in-index matching has an optimal time complexity, relatively expensive random memory access is required for elementary operations on the FM-index. We analyze to what extent in-index matching can be complemented with in-text verification where a candidate occurrence is directly validated in the text using a bit-parallel, pairwise alignment procedure. We find that hybrid in-index/in-text matching can reduce the running time by more than a factor of two, compared to pure in-index matching. We present Columba 1.1, an open-source (AGPL-3.0 license) software tool written in C++ that efficiently implements these ideas. Using a single CPU core, Columba 1.1 can identify, within a maximum edit distance of four, all occurrences of 100 000 Illumina reads (150 bp) in the human reference genome in roughly half a minute. This significantly outperforms existing, state-of-the-art tools