Three-dimensional modeling of chromatin structure from interaction frequency data using Markov chain Monte Carlo sampling

BMC Bioinformatics

Research Article

James Fraser¹  Josée Dostie¹  Maria A Ferraiuolo¹  Mathieu Rousseau²  Mathieu Blanchette² 

[1] Department of Biochemistry, and Goodman Cancer Research Centre, McGill University, 3655 Promenade Sir-William-Osler, Montréal, Québec, H3G 1Y6, Canada;McGill Centre for Bioinformatics, Bellini Building, Life Sciences Complex, 3649 Promenade Sir William Osler, H3G 0B1, Montréal, Québec, Canada;

关键词: Markov Chain Monte Carlo;    Base Density;    Interaction Frequency;    Markov Chain Monte Carlo Sampling;    HoxA Cluster;   

DOI  :  10.1186/1471-2105-12-414

 received in 2011-05-18, accepted in 2011-10-25,  发布年份 2011

来源: Springer

【 摘 要 】

BackgroundLong-range interactions between regulatory DNA elements such as enhancers, insulators and promoters play an important role in regulating transcription. As chromatin contacts have been found throughout the human genome and in different cell types, spatial transcriptional control is now viewed as a general mechanism of gene expression regulation. Chromosome Conformation Capture Carbon Copy (5C) and its variant Hi-C are techniques used to measure the interaction frequency (IF) between specific regions of the genome. Our goal is to use the IF data generated by these experiments to computationally model and analyze three-dimensional chromatin organization.ResultsWe formulate a probabilistic model linking 5C/Hi-C data to physical distances and describe a Markov chain Monte Carlo (MCMC) approach called MCMC5C to generate a representative sample from the posterior distribution over structures from IF data. Structures produced from parallel MCMC runs on the same dataset demonstrate that our MCMC method mixes quickly and is able to sample from the posterior distribution of structures and find subclasses of structures. Structural properties (base looping, condensation, and local density) were defined and their distribution measured across the ensembles of structures generated. We applied these methods to a biological model of human myelomonocyte cellular differentiation and identified distinct chromatin conformation signatures (CCSs) corresponding to each of the cellular states. We also demonstrate the ability of our method to run on Hi-C data and produce a model of human chromosome 14 at 1Mb resolution that is consistent with previously observed structural properties as measured by 3D-FISH.ConclusionsWe believe that tools like MCMC5C are essential for the reliable analysis of data from the 3C-derived techniques such as 5C and Hi-C. By integrating complex, high-dimensional and noisy datasets into an easy to interpret ensemble of three-dimensional conformations, MCMC5C allows researchers to reliably interpret the result of their assay and contrast conformations under different conditions.Availabilityhttp://Dostielab.biochem.mcgill.ca

【 授权许可】

Unknown   
© Rousseau et al; licensee BioMed Central Ltd. 2011. This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.