Matches in UGent Biblio for { ?s ?p Background: Over the last decade, DNA-methylation research has shifted from a gene-based approach to genome-wide analyses. DNA-methylation, featured by the enzymatic methylation of cytosines in a predominantly CpG-dinucleotide context, is an epigenetic process that is tightly associated with gene expression regulation. A novel generation of methodologies has enabled researchers to profile DNAmethylation in a genome-wide manner, e.g. by Methyl-Binding Domain (MBD)-based affinity purification followed by NGS (MBD-seq). While MBD-seq provides an excellent combination of sensitivity and cost-efficiency, it is featured by a set of bioinformatics and statistical challenges that complicates the subsequent data-analysis. The data-analysis pipeline for quantitative NGS applications typically consists of quality control, sequence mapping, data summary, data normalization and statistical analysis. Several of these steps require specific solutions for MBD-seq data: - Quality control is difficult yet necessary as sensitivity and specificity of enrichment procedures may vary. - Data summary is complicated by the lack of a functional unit for DNA-methylation, cf. the exon as unit for RNA-seq data summary. - Most data normalization procedures assume that the overall profiles are similar between samples, an assumption that is invalid for DNA-methylation. - The identification of significant enrichment is usually based on a Poisson background model. This model has several restrictions, resulting in suboptimal power. Therefore, we aimed at developing tailor-made solutions for each of these challenges. Methods: Captured fragments were paired-end sequenced (Illumina GAIIx). Sequenced reads were mapped on the human genome with BOWTIE. R, Perl, Java and MySQL were used to implement the different solutions. Results: We could demonstrate that the CpG-density profile of the sequenced fragments provides a solid basis for quality control, including the evaluation of sensitivity and specificity. A Map of the Human Methylome was constructed based on a large collection of MBD-seq profiles. This map consists of putatively independently methylated genomic regions, i.e. Methylation Cores (MCs), that can be used for data summary. For data normalization, a procedure called “Massively Enriched Loci Normalization" (MELON) was developed, based on the assumption that there exists a set of massively enriched loci of which the degree of DNA-methylation is similar between samples. A novel statistical framework, that provides higher power and sensitivity than the standard Poisson model has (partially) been developed, and can also be used for other enrichment based NGS applications. Conclusions: An optimized pipeline for high-quality MBD-seq data-analysis has largely been developed and implemented.. }
Showing items 1 to 1 of
1
with 100 items per page.
- aggregation abstract "Background: Over the last decade, DNA-methylation research has shifted from a gene-based approach to genome-wide analyses. DNA-methylation, featured by the enzymatic methylation of cytosines in a predominantly CpG-dinucleotide context, is an epigenetic process that is tightly associated with gene expression regulation. A novel generation of methodologies has enabled researchers to profile DNAmethylation in a genome-wide manner, e.g. by Methyl-Binding Domain (MBD)-based affinity purification followed by NGS (MBD-seq). While MBD-seq provides an excellent combination of sensitivity and cost-efficiency, it is featured by a set of bioinformatics and statistical challenges that complicates the subsequent data-analysis. The data-analysis pipeline for quantitative NGS applications typically consists of quality control, sequence mapping, data summary, data normalization and statistical analysis. Several of these steps require specific solutions for MBD-seq data: - Quality control is difficult yet necessary as sensitivity and specificity of enrichment procedures may vary. - Data summary is complicated by the lack of a functional unit for DNA-methylation, cf. the exon as unit for RNA-seq data summary. - Most data normalization procedures assume that the overall profiles are similar between samples, an assumption that is invalid for DNA-methylation. - The identification of significant enrichment is usually based on a Poisson background model. This model has several restrictions, resulting in suboptimal power. Therefore, we aimed at developing tailor-made solutions for each of these challenges. Methods: Captured fragments were paired-end sequenced (Illumina GAIIx). Sequenced reads were mapped on the human genome with BOWTIE. R, Perl, Java and MySQL were used to implement the different solutions. Results: We could demonstrate that the CpG-density profile of the sequenced fragments provides a solid basis for quality control, including the evaluation of sensitivity and specificity. A Map of the Human Methylome was constructed based on a large collection of MBD-seq profiles. This map consists of putatively independently methylated genomic regions, i.e. Methylation Cores (MCs), that can be used for data summary. For data normalization, a procedure called “Massively Enriched Loci Normalization" (MELON) was developed, based on the assumption that there exists a set of massively enriched loci of which the degree of DNA-methylation is similar between samples. A novel statistical framework, that provides higher power and sensitivity than the standard Poisson model has (partially) been developed, and can also be used for other enrichment based NGS applications. Conclusions: An optimized pipeline for high-quality MBD-seq data-analysis has largely been developed and implemented.".