UCL Cancer Institute

Analysis Pipelines

Project Leaders: Dr Gareth Wilson & Dr Tiffany Morris

We utilize and develop computational tools for the analysis of biological data, primarily obtained from studies of DNA methylation. The data predominantly comes from either second generation sequencing platforms or methylation arrays.

DNA Sequence Analysis

Our focus is on the analysis of MeDIP-seq, though we are also experienced in the analysis of RNA-seq, ChIP-seq, bis-seq and exome sequencing.

Originally, the Batman algorithm (Down et al., 2008) was used for our MeDIP analyses, including the first cancer methylome (Feber et al., 2011). Since then we have developed the MeDUSA pipeline (Wilson et al., submitted). The focus of which is to locate differentially methylated regions between cohorts.

MeDUSA (Methylated DNA Utility for Sequence Analysis) brings together numerous software packages to perform a full analysis of MeDIP-seq data, including sequence alignment, quality control (QC), and determination and annotation of DMRs. MeDUSA utilizes several applications from within the USeq software suite, and in turn uses the R Bioconductor package DESeq for differential count analysis. In addition, MeDUSA will control several other important functions from the alignment (BWA) and subsequent filtering (SAMtools), through generation of numerous quality control metrics (FastQC and MEDIPS), and finally preliminary annotation of the DMRs (utilizing the capabilities of BEDTools). MeDUSA can be downloaded from HERE.

A focus for future research within the group is on the integration of disparate datasets in order to elucidate a fuller understanding of the underlying biology and thus address fundamental questions associated with epigenetic regulation of mammalian cells.

Array Analysis

Another focus in our group is on analysis of Illumina's Infinium 450k methylation array. This platform was designed with two different probe types. Technical differences have been shown to exist between the two probe types and normalisation methods for dealing with that are being developed (Dedeurwaerder et al., 2011, Makismovic et al., in review). In addition, careful study design is important to avoid issues with batch effects.

Our group has assembled a pipeline that utilises the minfi R library to process the raw data files. The data is then normalised separately for each probe type and experimental and clinical information is incorporated to consider batch effects and potential effects from array position (Teschendorff et al, 2011).

Recently we hosted a workshop that dealt specifically with 450k analysis issues. Details and slides from this workshop are available HERE

References

1. Down TA, Rakyan VK, Turner DJ, Flicek P, Li H, Kulesha E, Graf S, Johnson N, Herrero J, Tomazou EM, et al: A Bayesian deconvolution strategy for immunoprecipitation-based DNA methylome analysis. Nat Biotechnol 2008, 26:779-785.
2. Feber A, Wilson GA, Zhang L, Presneau N, Idowu B, Down TA, Rakyan VK, Noon LA, Lloyd AC, Stupka E, et al: Comparative methylome analysis of benign and malignant peripheral nerve sheath tumors. Genome Res 2011, 21:515-524.
3. Sarah Dedeurwaerder, Matthieu Defrance, Emilie Calonne, Helene Denis, Christos Sotiriou, and Francois Fuks: Evaluation of the Infinium Methylation 450K technology Epigenomics 2011, 3:771-784.
4. Jovana Makismovic, Lavinia Gordon, and Alicia Oshlack: Subset quantile within-array normalization for Illumina Infinium HumanMethylation450 BeadChips, in review.
5. Teschendorff A E et al.: Independent surrogate variable analysis to deconvolve confounding factors in large-scale microarray profiling studies. Bioinformatics 2011;27:1496-1505