Detection of natural antisense transcripts conserved between human and mouse

Par Engstrom1, Hidenori Kiyosawa2, Claes Wahlestedt, Yoshihide Hayashizaki, Boris Lenhard, Center for Genomics and Bioinformatics, Karolinska Institutet, Stockholm, Sweden;, RIKEN Bio Resource Center, RIKEN Tsukuba Institute, Tsukuba, Japan

The mapping of cDNA and EST sequences to human and mouse genome assemblies has revealed thousands of pairs of overlapping and oppositely directed transcription units. In prokaryotes, naturally occurring complementary transcripts have been found to regulate expression of their corresponding gene(s). It has been demonstrated that an order of magnitude more of the human genomic sequence may be transcribed than accounted for by the predicted or characterised exons. Widespread antisense transcription in human has also recently been shown experimentally. However, the experimental support for antisense activity and biological role of eukaryotic transcripts is still scarce. Starting from the hypothesis that functionally important antisense transcripts are more likely to be conserved between human and mouse, we assessed the extent of conservation for 1964 pairs of overlapping mouse transcription units predicted by cDNA mapping (Kiyosawa et al., Genome Res. in press). We retrieved human homologs of the mouse transcripts, and designed an automated computational procedure for the detection of human sense-antisense pairs equivalent to those in mouse. Genomic regions were located and gene structures predicted using cDNA mappings from the the UCSC Genome Browser Database, or direct mapping to the corresponding genome assembly using BLAT. Homologous human and mouse genomic regions were aligned to determine whether the region of putative sense-antisense overlap in the mouse genome corresponded to a region of sense-antisense overlap in the human genome. We divided transcribed region overlaps into five classes based on the type of overlap, as originally done by Kiyosawa et al., and showed that the level of cross-species support varies significantly based on the type of overlap. A subset of the gene pairs identified by this approach are currently being screened for antisense activity by laboratory methods.