Short Abstract: Several high-throughput methods have been developed to detect DNA methylation on a genomic scale. Among these, Illumina’s HumanMethylation450 (HM450 bead array) BeadChip assesses the methylation status of over 450,000 CpG positions in the human genome. The technology relies on hybridization of genomic fragments to probes on the chip. A two-colour channel array design provides a signal based on the methylation status of a CpG. However, it has been previously noted that factors other than methylation changes can alter the inferred methylation status. These include single nucleotide polymorphisms (SNPs), small insertions and deletions (INDELS) and repetitive regions of DNA. Prior to hybridization, the genome is bisulfite treated, which results in many probes on the array no longer mapping to unique locations. These factors have the potential to give rise to false methylation calls. Currently, there is no clear method or pipeline for determining which of the probes on the HM450 bead array should be retained for subsequent analysis. Exclusion of these affected probes in the data processing procedure can substantially improve the methylation estimates.
We propose a method that enables identification of the affected probes on the HM450 bead array. We use additional information such as SNP status to ensure only the minimal probe set is discarded. Our method significantly reduces the risk of false discoveries while maximising the power of the HM450 bead array to detect methylation status genome-wide.

Short Abstract: Bisulfite conversion followed by next generation sequencing or microarray-based profiling (e.g. using Illumina’s HumanMethylation450 platform) have become popular means of assessing genome-scale DNA methylation profiles. They are routinely employed in epigenome-wide association studies and large-scale epigenomic analysis projects.
We developed RnBeads, an R package implementing a comprehensive and user-friendly pipeline for DNA methylation analysis and interpretation. A large variety of input formats are supported. The pipeline implements state of the art normalization techniques. Experimental quality control can be conducted and sample outliers and mix-ups can be identified. The package allows for CpG and sample filtering due to a multitude of criteria. According to sample annotations, batch effects and phenotype covariates can be identified. DNA methylation distributions are analyzed and intergroup as well as intragroup variability in methylation profiles is quantified. Furthermore, differential methylation between groups of samples can be characterized. The analysis is based on individual CpGs as well as on predefined or custom genomic regions. Finally, methylation data can be exported in various formats including genome browser views.
Comprehensive, highly interpretable reports containing method descriptions, publication grade plots and data tables are generated. Their HTML format facilitates easy tracking and comparison of analyses as well as exchanging results with collaboration partners. Due to its modularized concept, both first-time users and experts can conveniently perform analyses according to their individual demands. A single, comprehensive analysis run can be invoked by specifying only few parameters and executing a master command. Alternatively, a user may execute the steps of the pipeline individually.

Short Abstract: Targeted bisulfite sequencing is a cost-efficient method for DNA methylation profiling at single-nucleotide resolution. After sequencing and data preprocessing, the number of methylated and unmethylated reads are obtained for each covered CpG-site. So far, only few tools to process and analyze this kind of data exist.

We implemented the R/Bioconductor package BiSeq, which provides useful classes and functions to handle, visualize and analyze targeted bisulfite sequencing data. In particular, it implements an algorithm to detect differentially methylated regions (DMRs) between two groups of samples.

The DMR detection is divided into five steps: 1.) Find target regions. 2.) Smooth methylation levels in these target regions for each sample. 3.) Estimate the group effect along the CpG-sites via beta regression. 4.) Test for significant group effects in each region and control a region-wise FDR. 5.) Test single CpG-sites within significant regions and control a location-wise FDR.

In addition to the methods for DMR detection there are a couple of methods for quality control and to manipulate or filter bisulfite sequencing data. We also implemented methods to visualize methylation levels across genomic regions and samples as well as methods for clustering.

Short Abstract: Genome-wide functional assays based on high-throughput sequencing now allow for experimental probing of a wide variety of molecular phenotypes. Among these is DNA methylation, which can be probed at all CpG sites in the genome using bisulfite sequencing. This has allowed for comparisons of methylation extent in different functional regions by first averaging methylation states within region types and then comparing averages between regions. Such comparisons have become commonplace in genome-wide DNA methylation studies. For example, it has been repeatedly reported that the methylation extent is significantly higher in coding regions as compared to introns or UTRs. We report and characterize a bias present in these seemingly straightforward comparisons that is a special case of the Yule-Simpson's effect and show it has extensively altered the magnitude and significance of DNA methylation differences observed and reported from such comparative studies. The bias we discuss arises from the dependance of the sparsity of CpG sites on the extent of evolutionary pressure at a region, together with its overall methylation state. We present a correction utilizing a matrix completion algorithm that is based on a methylation model and show how it affects reported results regarding differences in DNA methylation across functional regions.

Short Abstract: Expression of genes is regulated by different factors. Epigenetic changes such as histone modifications and chromatin remodeling regulate the gene expression during skeletal muscle development. The aim of our study was to identify the role of chromatin remodeling factors and histone marks in regulating myogenesis. We performed chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) to identify genome-wide binding of three different components of the BAF chromatin remodeling complex (Dpf3, Baf60c, Brg1) and three different histone marks (H3ac, H3K4me2, H3K4me3) in undifferentiated C2C12 cells (myoblasts) as well as differentiated C2C12 cells (myotubes). In addition to ChIP-seq, we gathered mRNA profiles (RNA-seq) from both cell types.

To identify the target genes of the BAF complex components and histone marks in both myoblasts and myotubes as well as differentially expressed genes during differentiation, an analysis pipeline was established utilizing ChIP-seq and RNA-seq data. The pipeline includes several quality control steps (e.g. using FastQC after sequencing), statistical reporting, mapping of sequencing reads (Bowtie), ChIP-seq peak calling (MACS), peak visualization in IGV or UCSC, and differential expression analysis (Cufflinks).

In summary, we identified different epigenetic regulation between myoblasts and myotubes. Moreover, the developed analysis workflow is not only limited to myogenesis. It can be used to analyze the differentiation of other cell types (e.g. fibroblasts and fibrocytes) from ChIP-seq and RNA-seq datasets.

Short Abstract: ChIP-seq, chromatin immunoprecipitation followed by high-throughput sequencing, is the technique of choice for accurate characterization of transcription factor binding sites (TFBSs) and detection of histone modifications (HMs). ChIP-seq is frequently used to study gene regulation based on epigenetic differences or differences in TFBSs between specific cellular conditions.

There exist many computational tools for ChIP-Seq data analysis. Most of them are command line applications or R packages and may be difficult to use for non-bioinformatician users. Thus, we created a web service, Nebula, which allows inexperienced users to perform a complete bioinformatics analysis of ChIP-seq data [1], http://nebula.curie.fr/.

The original pipeline [1] was designed to detect and annotate TFBSs. Here, we extended Nebula’s functionalities to be able to analyze HM data. The analysis workflow allows read mapping, peak calling and annotation of peaks with genomic features. For each RefSeq gene, Nebula reports detected HMs in the vicinity of the gene transcription start site (TSS) or in the gene body. The workflow accepts user-provided gene annotations such as gene expression data. Using user-provided partitioning of genes into high-/low-expressed or activated/inhibited/non-modulated by some factor, Nebula will visualize distribution of the signal around gene TSS separately for each gene group.

Nebula accepts both unmapped reads (Illumina FASTQ or SOLiD format) and mapped reads (SAM/BAM format). Each step of the pipeline produces several standardized output files (such as WIG and BED) which can be used by third-party tools.

1. Boeva V, Lermine A, et al. (2012). Bioinformatics 28:2517–2519.

Short Abstract: A variety of post-­‐translational modifications, including
phosphorylation, acetylation, ubiquitylation, methylation and SUMOylation
(Kouzarides 2007; Bhaumik et al., 2007, Shilatifard 2006) occur on the amino
terminal tail, as well as on the residues located at exposed sites within the
globular domain of the histones (Couture and Trievel, 2006). Such modifications
on histones may play different roles. They can create or stabilize binding
sites for regulatory proteins such as transcription factors, proteins involved
in chromatin remodeling or DNA repair. Disruption or masking of
chromatin-­‐binding sites has also been reported as an effect of histone
modifications (Lee et al., 2010). The regulation of different biological
processes such as transcription, DNA repair and recombination and RNA
processing relies on an implementation/removal of histone tail modifications
(Bhaumik et al., 2007, Shilatifard 2006).
Two prominent protein groups, Trithorax and Polycomb group catalyse H3K4 and
H3K27 tri-methylation respectively. Trimethylation of H3K4 is associated with
transcriptional activation. In mammalian cells there are six different H3K4
methyltransferases which associate in high molecular weight complexes. Despite
they have been partly or fully biochemically characterized, their function is
still largely unknown. Major questions about these protein complexes have not
been answered yet. Do they have specialized roles? Could they be redundant and
compensate each other? If yes, who shares the more workload?
Moving on with these questions, we have acquired ChIP-Sequencing datasets and a
fully automated computational pipeline has also been developed to analyse the
plethora of high throughput datasets. We have also discovered a novel Setd1a
complex member which highly correlates with the methyltransferase and its
subunits.

Short Abstract: H3K27me3 is a histone modification that is correlated with polycomb mediated transcriptional silencing of genomic regions. H3K27me3 occurs in large blocks which renders the application of classical ChIP-seq peak finding algorithms inappropriate. Here we developed a novel HMM based approach that allows for the identification of large regions that carry the H3K27me3 histone modification, and more importantly, also for the identification of differentially modified regions between samples. We evaluated the performance of this approach using simulations as well as comparisons between biological replicates and comparisons to gene expression data. Then we analyzed two rat inbred strains BN.Lx and SHR.Ola to identify regions that are differentially modified. We found that differential regions were enriched for differentially expressed genes. In addition we experimentally validated differential calls for a randomly selected subset of regions.

Short Abstract: DNA methylation plays an important role for epigenetic gene regulation in development and disease. A CpG dinucleotide (CpG site) is called methylated, if a methyl group is attached to its C base. Bisulfite treatment followed by next generation sequencing is a method for DNA methylation profiling. After sequence alignment, the number of methylated and the number of unmethylated reads are obtained for each covered CpG site.

To detect methylation differences between two groups of samples, often, each CpG site is tested separately. Depending on the experimental setup, millions of CpG sites can be covered leading to a massive multiple testing problem. If genomic regions of interest can be specified, e.g. promoter regions, the number of tests can be reduced by testing each region instead of single CpG sites. A meaningful null hypothesis for regions is: “no CpG site within the region is differentially methylated”, which is denoted as self-contained null hypothesis in the context of gene set testing (Goeman and Bühlmann, Bioinformatics, 2007).

We compared three methods for testing genomic regions: (i) Poisson regression model, (ii) Z-test as used by Hebestreit et al. (in review) in a hierachical two step testing approach and (iii) the global test proposed by Goeman et al. (Bioinformatics, 2004) for gene expression data. Methylation differences were simulated and incorporated into a real bisulfite sequencing data set. Then, ROC analyses were performed based on this data set. The global test showed to be superior followed by the Z-test, which performed better than testing within a Poisson regression model.

Short Abstract: Promoters are the starting site of transcription and serve as binding site for RNA polymerases. In these regions different regulatory inputs from cis-regulatory elements are integrated. Promoters can be classified into CpG enriched (HCP) and CpG depleted (LCP). Here, we subclassified promoters by Principal Component Analysis (PCA) of expression patterns of ubiquitously and differentially expressed genes among 16 human tissues. Then, we analyzed DNA accessibility by DNaseI-seq and also histone modification patterns, which were shown to correlate quantitatively with expression of downstream genes [1]. Our results suggest that differentially expressed genes are driven by LCPs and are mainly regulated at the stage of transcription initiation. Developmental regulator genes, which are driven by HCPs, are likely to be regulated in the same way. In contrast, the majority of HCPs drive genes, which encode for essential transcripts of cell metabolism and might be regulated after transcription initiation at the level of transcription elongation.

[1] Rosa Karlic, Ho-Ryun Chung, Julia Lasserre, Martin Vingron et al.: Histone modification levels are predictive for gene expression, PNAS 2010

Short Abstract: The recent discovery that mammalian mitochondria contain an enzyme with the potential to methylate cytosine bases within the mitochondrial genome suggests a potential mechanism for the epigenetic regulation of mitochondrial function. However, the promising and emerging field of mitochondrial epigenetics is in its infancy and evidence for this epigenetic mark in mitochondrial DNA is still controversial.

To determine the methylation at base resolution, we focused on the analysis of mitochondrial DNA methylation in mouse muscle cells by using bisulfite sequencing. Two biological replicates of muscle samples were enriched for mitochondria and DNA extracted. DNA samples were then treated with sodium bisulfite, which allows the measurement of cytosine methylation at single-base resolution. To obtain highly accurate sequence reads we used the ligation-based 5500 SOLiD System with the additional Exact Call Chemistry module. Recent developments in software to align SOLiD-specific color space bisulfite sequencing data allowed comparison of different alignment strategies. Additionally, unconverted DNA was also sequenced to incorporate the underlying genetic variability of mouse mitochondrial DNA into our DNA methylation analysis.

On a genomic scale we identified low average methylation levels in both muscle samples. Additionally, consistent higher methylation was detected at individual sites. Efforts to experimentally confirm local patterns of higher methylation by methods such as hairpin-bisulfite PCR are not consistent. One possible explanation for this could be that methylation events only occur in a certain subset of the multiple copies of mitochondrial genomes within cells, which makes it difficult to consistently detect methylation without high throughput approaches.

Short Abstract: Development of tools to analyze and visualize the epigenome remains a challenge. chroGPS is a computational approach that addresses this question. ChroGPS represents epigenetic factors or genetic elements based on their epigenetic state in 2D/3D reference maps, allowing the integration of large amounts of genetic and epigenetic data from heterogeneous sources. The result is an easy-to-interpret representation of the relationships between the factors/elements that includes relevant information about their functional properties. chroGPS is a general methodology that can be applied to study the epigenetic state of any class of genetic element or genomic region. To illustrate the usefulness of these maps to interpret epigenetic information in a functional context and derive testable hypotheses, we present several case studies using data on the genomic distribution of a large collection of chromatin components generated by the modENCODE project in Drosophila melanogaster, a model system extensively used to study genome function.

Short Abstract: Chromatin accessibility is a predominant feature of regulatory elements in the genomes of eukaryotes. We investigate the location and function of regions of cell-type-specific chromatin accessibility in the human genome, using DNase-Seq data from nine fetal and two non-fetal primary tissues as well as two cell lines, provided by the Roadmap Epigenomics Mapping Consortium. By taking into account the height of DNase-Seq signal as well as the variability of the signal across all samples from all tissues, we are able to not only identify but quantify cell-type-specificity of chromatin accessibility. Top ranked cell-type-specific DNase-hypersensitive sites (CTS-DHS) are enriched with motifs of developmental transcription factors for the relevant tissue and are located near genes with cell-type-specific gene expression. The importance of these regulatory regions to the tissues is further underscored by a significant overlap of SNPs associated with diseases or traits in relevant organ systems. Thus, the cell-type-specific regulatory regions we identify help to characterize those mechanisms by which genetic variants produce disease phenotypes.

Short Abstract: DNA methylation is an epigenetic mark that has a regulatory role in a broad range of biological processes and diseases. It is particularly important in cancer cells, where aberrant DNA methylation profiles are extremely frequent and regarded as a potential biomarker for the diagnosis and treatment of the disease. The downstream effects of these abnormal profiles on gene expression are, however, poorly understood. In this work, we analyze the transcriptome and DNA methylation patterns of two types of pediatric brain tumor samples displaying distinct expression levels of the known DNA methylases. We present an integrative analysis, characterizing the distribution of methylated sites across the genome and its relationship with particular features, such as CGI islands, transcription factor binding sites and histone modification sites. By systematically partitioning the regions surrounding transcription start sites and analyzing the correlation of methylation and expression estimates, we are able to identify at high resolution the genomic regions predictive of gene expression in these tumor types.

Short Abstract: Multipotent CD4+ T cells are central to the adaptive immune system. CD4+ T cells can differentiate to functionally distinct effector subtypes such as T helper 1 (Th1), Th2, Th17, and iTreg. In this study, we have focused on identification of histone modifications (H3K4me1, H3K27ac, H3K4me3) that define the cell-type specific functional cis-regulatory repertoire for early differentiating human Th1 and Th2 cells. Additionally, we have integrated genome-wide digital gene expression analysis from the Helicos platform to correlate epigenetic information with gene expression. We also overlay the identified enhancer regions with open chromatin sites (DNase-seq) from fully differentiated T cells to characterize whether early enhancers are active only during the early lineage specification or remain active in committed Th cells. By analyzing transcription factor binding sites at enhancers we are able to identify known and novel transcriptional regulators which drive the lineage determination. Lastly, under the principle that improper cell fate specification can lead to immunopathogenesis, we found within these lineage-specific enhancers a great number of SNPs from genome-wide association studies (GWAS) that were associated with various autoimmune disorders including T1D, rheumatoid arthritis, Crohn’s disease, and asthma. Several alter transcription factor binding site motifs, and using DAPA experiments we show for a subset of such SNPs within these predicted sites that they influence transcription factor binding. This study provides the first look at how enhancers can contribute to early human T cell lineage specification. Our results also provide insight into how regulatory SNPs may contribute to the disease pathogenesis.

Short Abstract: DNA methylation is an epigenetic modification that predominantly affects cytosine nucleotides in CpG context without changing the DNA sequence. It is generally expected that the maternal and the paternal copies of the human genome display similar methylation patterns within cells of the same tissue. However, genetic differences might result in differential methylation of parental DNA copies. Furthermore, biological processes such as inactivation of one female X-Chromosome, and parental imprinting of genes are causes of allele-specific epigenetic DNA methylation patterns. Via bisulfite-conversions DNA methylation can be measured by next-generation sequencing technology. Around single nucleotide polymorphisms (SNPs) this information can be analyzed for allele-specificity, while in the remaining genome the signals of both alleles are convoluted. For 12 whole-genome methylomes, we analyzed the amount of obtainable information, and found that depending on the sequencing depth and SNP location 0.1-2.7 % of all CpGs can be analyzed for allele-specific methylation patterns. Of these, 14-21% display evidence for allele-specific methylation.
We here present ASMHunter, a software pipeline, which guides this analysis workflow from the primary sequencing data to the visualization of the results as comprehensive genome browser tracks.

Short Abstract: Driven by the advance of high throughput sequencing (HTS) technology, several new technologies for whole (epi)genome analysis have been developed in recent years. Thanks to these methods, genetic variants, gene expression, and DNA methylation is now reliably detectable at a genome wide scale. Alteration of DNA methylation is a hallmark of cancer. It is thought to locally influence the regulatory potential of DNA, leading to changes in gene expression during development and disease. This correlation has been shown for selective cases, but general relations are not yet known.
Here, we present a computational work flow, MEDIPS, for the analysis of HTS methylome data, and strategies to integrate the results with transcriptome and genome variation data sets. We assess the correlation of methylation, copy-number variations and differential gene expression in CpG Island and non-CpG Island promotors, transcription factor binding sites, and other annotated regulatory features in mouse and human colon cancer data.
We identify focal differential methylation as the cause of gene expression de-regulation. Furthermore, we show that genomic copy-number variations influence both methylation and gene expression. This interdependency is shown with selected examples, as well as genome wide correlation.

Short Abstract: Hi-C is a new experimental method to determine the three-dimensional structure of a genome. The technique involves fixing chromatin with formaldehyde to preserve its structural integrity, followed by restriction enzyme digestion, ligation and then sonication to generate a population of short DNA sequences termed di-tags which reflect the spatial arrangement of the genome at the time of fixation.

Numerous statistical analyses are being developed to translate the population of sequenced di-tags into three-dimensional models of the genome, providing insights into spatial organisation and how chromosome folding may affect gene expression. While complex analytic techniques are necessary to build robust three-dimensional structures, the predictive power of such methods is limited by the quality of the input data.

A number of steps in the Hi-C protocol have been identified that can lead to biases or noise in the sequenced di-tag population. Fortunately, many of these experimental artefacts can be identified and removed from the final dataset prior to statistical analysis. Here we present our examination of Hi-C artefacts generated during the experimental process, showing why it is import to filter them out and how our Hi-C pipeline, HiCUP, has improved Hi-C datasets.