Biclustering of Expression Data
Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology
| Hi-index | 0.00 |
This paper presents a method for predicting biologically meaningful modules of transcription factors. For this purpose, we employ the CORG database of conserved transcription factor binding sites. We aim at enhancing the power of in-silico binding site predictions by employing three crucial constraints. First, we rely on conserved promoter regions of orthologous genes in human and mouse, second we look for synergistic transcription factor modules which bind upstream regions preferentially together, and finally we restrict our results to those modules, whose genes have a significant functional overlap. Many of our predicted binding sites coincide with known biological facts as is evidenced by a direct comparison with a single large-scale experiment for E2F binding. We also identified known combinations of transcription factors with a functional enrichment in the set of their shared target genes. Several new modules are suggested for experimental investigation. Finally we study the transcription factor network and suggest a classification of transcription factors according to their regulatory power and control.