Prediction of Cis-Regulatory Elements of Coregulated Genes

Szymon M. Kielbasa[1] (s.kielbasa@itb.biologie.hu-berlin.de)
Nils Blüthgen[1] (nils@itb.biologie.hu-berlin.de)
Christine Sers[2] (christine.sers@charite.de)
Reinhold Schäfer[2] (reinhold.schaefer@charite.de)
Hanspeter Herzel[1] (h.herzel@biologie.hu-berlin.de)

[1]Institute for Theoretical Biology, Humboldt University, Berlin, Germany
[2]Institute of Pathology, Charité, Humboldt University, Berlin, Germany

Abstract

We present a computational pipeline to predict cis-regulatory elements composing results based on different algorithms: Clover, Cluster-Buster, an own implementation of human/rat/mouse sequence identity and our ITB algorithm. The procedure uses information from the human genome sequence, NCBI gene annotations, verified eukaryotic promoters (EPD), experimentally proven binding sites (Transfac) and homologies to mouse and rat (HomGL/HomoloGene).
  We test the approach on 18 upstream regions of experimentally verified AP-1 target genes. About a half of the known sites belong to high-scoring candidates. Three top-scoring elements are confirmed by Cluster-Buster and high homologies.
  The same analysis we applied to genes found to be up- or downregulated due to mutated RAS. We performed a detailed literature and computational search for promoter regions. Indications of overrepresented Elk-1 and AP-1 motifs are found via a comparison with shuffled sequences. In some promoters consistent predictions of clustered binding sites were obtained.

[ Full-text PDF | Table of Contents ]