Distinguishing between genomic regions bound by paralogous transcription factors

Authors:
Alina Munteanu;Raluca Gordân
Affiliations:
Faculty of Computer Science, Alexandru I. Cuza University, Iasi, Romania;Institute for Genome Sciences and Policy, Departments of Biostatistics & Bioinformatics, Computer Science, and Molecular Genetics and Microbiology, Duke University, Durham, NC
Venue:
RECOMB'13 Proceedings of the 17th international conference on Research in Computational Molecular Biology
Year:
2013

Citing 5
Cited 0

Support-Vector Networks

Machine Learning
Random Forests

Machine Learning
On safari to Random Jungle

Bioinformatics
LIBSVM: A library for support vector machines

ACM Transactions on Intelligent Systems and Technology (TIST)
MEME-ChIP

Bioinformatics

Quantified Score

Hi-index	0.00

Visualization

Abstract

Transcription factors (TFs) regulate gene expression by binding to specific DNA sites in cis regulatory regions of genes. Most eukaryotic TFs are members of protein families that share a common DNA binding domain and often recognize highly similar DNA sequences. Currently, it is not well understood why closely related TFs are able to bind different genomic regions in vivo, despite having the potential to interact with the same DNA sites. Here, we use the Myc/Max/Mad family as a model system to investigate whether interactions with additional proteins (co-factors) can explain why paralogous TFs with highly similar DNA binding preferences interact with different genomic sites in vivo. We use a classification approach to distinguish between targets of c-Myc versus Mad2, using features that reflect the DNA binding specificities of putative co-factors. When applied to c-Myc/Mad2 DNA binding data, our algorithm can distinguish between genomic regions bound uniquely by c-Myc versus Mad2 with 87% accuracy.