An oligo-to-genome alignment algorithm capable of exhaustively finding imperfect matches in a large genomic database even with very short oligos. Designed to aid prediction of potential off-target activity in Anti-Sense Oligos (ASOs), where biological activity had been noted even with two mismatches in an oligo aligning to an intronic region.
The need for this tool developed with the recognition (empirically demonstrated by bench experiments) that transcripts could be knocked down by imperfect alignment to intronic sequences. In addition, it was important to prioritize ASO candidates that showed reasonably homology to model organisms, in order to facilitate in vivo testing. Several open source and commercial tools were considered, but all failed to exhaustively identify partial-mismatch locations with small oligos.
An exhaustive description of both database construction and the sarch algorithm is available, but in broad strokes proceeds as:
The search database is a Postgres DB centered around polymorphism-permuted 12-mer oligos. The sequence source is the set of all "pre-mRNA" sequences for several genomes. One sequence is used for each gene, representing the full genomic sequence spanning from the "left-most" edge of all transcripts for that gene to the "right-most". This allows both intronic and exonic sequence to be considered. In addition, splice boundaries and ORFs for each variant are also recorded, which allows hits to be assigned to specific transcripts, as well as allowing classification of hits into exonic, CDS, splice acceptor, UTR etc.
Multiple genomic builds are included in the database. In our production, these represented a couple human builds (GRCh37 plus RefSeq Gene), several primates, dog, mouse and rat.
When an oligo (ASO) is submitted for searching, a minimum tiling of 12-mer words is generated for it. The user also specifies the number of mismatches they wish to allow, the default being 2. Each word is then fully permuted for all possible mismatch combinations. This permuted set is then searched in parallel (this is Perl, so forked) against the database. For speed, some of the permutations will be discarded if prior, neighboring oligo hits preclude the possibility of finding a full-oligo match within the requested mismatch limit.
Results are then provided in three forms:
- Multi-sheet Excel workbook
- Interactive DHTML tables
- Interactive DHTML genomic browser
An example report from the interactive table browser (showing published oligos from the literature):
