MSCAN: identification of functional clusters of transcription factor binding sites

Abstract

Identification of functional transcription factor binding sites in genomic sequences is notoriously difficult. The critical problem is the low specificity of predictions, which directly reflects the low target specificity of DNA binding proteins. To overcome the noise produced in predictions of individual binding sites, a new generation of algorithms achieves better predictive specificity by focusing on locally dense clusters of binding sites. MSCAN is a leading method for binding site cluster detection that determines the significance of observed sites while correcting for local compositional bias of sequences. The algorithm is highly flexible, applying any set of input binding models to the analysis of a user-specified sequence. From the user's perspective, a key feature of the system is that no reference data sets of regulatory sequences from co-regulated genes are required to train the algorithm. The output from MSCAN consists of an ordered list of sequence segments that contain potential regulatory modules. We have chosen the features in MSCAN such that sequence and matrix retrieval is highly facilitated, resulting in a web server that is intuitive to use. MSCAN is available at http://mscan.cgb.ki.se/cgi-bin/MSCAN.

Figure 1

Screenshot of the MSCAN data input page. The initial input form contains fields for the provision of TF binding profiles (matrices), sequences and parameters. Links to the help file provide users with information about formats and description of the parameters.

Figure 2

Screenshot of the output page. Shown is a page with a predicted module in the regulatory region of a gene known to be selectively expressed in a context linked to the selected TFs. Every category label on the output page is hyperlinked to the help page.