ASTRANOM System for TRAnscriptional Networks Observation and Modelling
By Eugene Berezikov
Dr., Hubrecht Laboratory
In silico prediction of transcriptional regulatory networks would be a valuable tool for the wet lab biologist to make hypothesis and design experiments to elucidate the real transcriptional interactions in an organism. The key step in building a transcriptional network in silico is prediction of target genes for a particular transcriptional factor (TF). Since searching a sequence with a pattern or position weight matrix often produces high rates of false positives, the preffered method is ?phylogenetic footprinting?, where orthologous promoters from different organisms are aligned and searched for TF binding sites. We have developed alternative so-called Site Driven Alignment (SDA) approach to detect conserved regulatory sequence in orthologous promoters. In SDA, orthologous sequences are first searched for binding sites of a set of transcriptional factors and all hits of respective factors are compared pairwise and sorted by percentage of homology. Next, hit pairs (starting with most homologous) are placed in an iterative manner into an alignment in such a way that newly placed hits doesn't overlap with already established alignment setpoints. It appeared that SDA approach produces same results as conventional phylogenetic footprinting on promoters where alignment is obvious and outperfoms in cases were establishment of correct alignment is more tricky due to, for example, incorrect promoter annotation. Hence we concluded that SDA is more sutaible for genome-wide automatic prediction of transcription factor binding sites.
We have utilized SDA approach to annotate 15242 orthologous human/mouse promoter pairs extracted from Ensembl. The search was performed using TFBS software (B. Lenhard and W.W.Wasserman) and 357 human/mouse matrices from TransFac Pro 6.3 database. Search results were stored in MySQL database and web interface called ATSRANOM was developed to access and visualize the data. ASTRANOM system (http://rat.niob.knaw.nl/astranom) allows user to search promoters by name, Ensembl id and GO terms, visualize predicted transcription factors in individual promoter pairs, lookup target genes for a particular transcription factor/matrix and finally, build transcriptional regulatory networks with certain parameters. It is also possible to make user defined pattern/matrix searches on the set of orthologous promoters and visualize results along with precalculated predictions.