APIII - Advancing Practice, Instruction & Innovation Through Informatics

Visualization and Analysis of Whole Genome Epigenetic Data

Devon Macey, (dmacey@bccrc.ca); Andrew Thomson, Anna Chu, Jonathan J. Davies, Raj Chari, and Wan L. Lam; British Columbia Cancer Research Centre, University of British Columbia, Vancouver, BC, Canada

Context:  Array comparative genomic hybridization (aCGH) is a widely used approach by cancer researchers to detect DNA copy number alterations.  An adaptation to aCGH to study genomic DNA methylation patterns has recently been developed.  Visualization and analysis of this data requires individual plotting and reassembly of each data point’s ratio into individual chromosome profiles for a whole genome view of methylation status.  As no standards exist for this developing field, a specialized software suite with the ability to analyze methylation data has been designed and implemented.

Technology:  Methylation-dependent immunoprecipitation (MeDIP) is a technique used for isolating methylated DNA fragments.  MeDIP, in conjunction with aCGH, uses a bacterial artificial chromosomes microarray contiguously spanning the human genome, allowing for the detection of regions of increased and decreased methylation.  SIGMA (System for Integrated Genomic Microarray Analysis) Viewer has been implemented in Java, using the database application MySQL and the statistical package R to compute various analysis algorithms.

Design:  SIGMA Viewer was built for the purpose of analyzing and viewing multiple platforms of high-throughput genetic data, including methylation data.  SIGMA Viewer provides a highly efficient analysis ‘pipeline’, from the input of raw data to the final statistical analyses involved in delineating regions of differential methylation.  Specifically, a rank based approach is used to normalize log2 signal rations to facilitate inter-experiment comparison.  For cancer samples, an adjustment for methylation levels in normal cells is also performed.  This built-in processing of whole genome methylation data allows for the use of segmentation algorithms that are normally employed for standard DNA copy number array CGH.  Finally, linkage of interesting regions of differential methylation to biological databases such as the UCSC Genome Browser and NCBI Entrez allows for quick access to important relevant information.  For a dataset, we utilized two previously published lung cancer cell lines with their matched blood lymphocyte lines to test our system.

Results:  Visualization and analysis using SIGMA Viewer has successfully replicated the previous analysis done on this dataset and identified regions of hyper- and hypomethylation.  Significantly, the amount of time to perform this pipeline using limited tools such as Microsoft Excel is drastically reduced when done through SIGMA Viewer.

Conclusion:  In conclusion, we have created an application to facilitate the visualization and analysis of whole genome DNA methylation data.  This software will be free to all academic researchers and will continually evolve to meet the demands of this new and emerging field of study.