The Genome Informatics group is headed by Prof. Sven Rahmann.
It is part of the Faculty of Medicine of the University of Duisburg-Essen and located at the University Hospital Essen.
Research and teaching are carried out in close cooperation with the chair for Algorithm Engineering (LS 11) of the Faculty of Computer Science at TU Dortmund.
The article “Analysis of min-hashing for variant tolerant DNA read mapping” by Jens Quedenfeld (now at TU Munich) and Sven Rahmann has received the Best Paper Award at the Workshop of Algorithms in Bioinformatics (WABI) 2017, held in Cambridge, MA, USA, August 20-23, 2017.
The authors consider an important question, as DNA read mapping has become a ubiquitous task in bioinformatics. New technologies provide ever longer DNA reads (several thousand basepairs), although at comparatively high error rates (up to 15%), and the reference genome is increasingly not considered as a simple string over ACGT anymore, but as a complex object containing known genetic variants in the population. Conventional indexes based on exact seed matches, in particular the suffix array based FM index, struggle with these changing conditions, so other methods are being considered, and one such alternative is locality sensitive hashing. Here we examine the question whether including single nucleotide polymorphisms (SNPs) in a min-hashing index is beneficial. The answer depends on the population frequency of the SNP, and we analyze several models (from simple to complex) that provide precise answers to this question under various assumptions. Our results also provide sensitivity and specificity values for min-hashing based read mappers and may be used to understand dependencies between the parameters of such methods. This article may provide a theoretical foundation for a new generation of read mappers.
The article can be freely accessed in the WABI conference proceedings (Proceedings of the 17th International Workshop on Algorithms in Bioinformatics (WABI 2017), Russell Schwartz and Knut Reinert (Eds.), LIPICS Vol. 88).
This work is part of subproject C1 of the collaborative research center SFB 876.
Christopher Schröder and Sven Rahmann
Algorithms for Molecular Biology
The beta distribution is a continuous probability distribution that takes values in the unit interval [0,1]. It has been used in several bioinformatics applications to model data that naturally takes values between 0 and 1, such as relative frequencies, probabilities, absolute correlation coefficients, or DNA methylation levels of CpG dinucleotides or longer genomic regions. One of the most prominent applications is the estimation of false discov ery rates (FDRs) from p-value distributions after multiple tests by fitting a beta-uniform mixture. By linear scaling, beta distributions can be used to model any quantity that takes values in a finite interval [L, U ]⊂R.
We show that the Maximum likelihood estimation for Beta distributions, MLE has significant disadvantages for beta distributions. The main problem is that the likelihood function is not finite (for almost all parameter values) if any of the observed data points are xi=0 or xi=1.
For mixture distributions, MLE frequently results in a non-concave problem with many local maxima, and one uses heuristics that return a local optimum from given starting parameters. Because already MLE for a single beta distribution is problematic, EM does not work for beta mixtures, unless ad-hoc corrections are made. We therefore propose a new algorithm for parameter estimation in beta mixtures that we call iterated method of moments.
Christopher Schröder*, Elsa Leitão*, Stefan Wallner, Gerd Schmitz, Ludger Klein-Hitpass, Anupam Sinha, Karl-Heinz Jöckel, Stefanie Heilmann-Heimbach, Per Hoffmann, Markus M. Nöthen, Michael Steffens, Peter Ebert, Sven Rahmann and Bernhard Horsthemke
* Contributed equally
Epigenetics & Chromatin 2017
There is increasing evidence for inter-individual methylation differences at CpG dinucleotides in the human genome, but the regional extent and function of these differences have not yet been studied in detail. For identifying regions of common methylation differences, we used whole genome bisulfite sequencing data of monocytes from five donors and a novel bioinformatic strategy.
We identified 157 differentially methylated regions (DMRs) with four or more CpGs, almost none of which has been described before. The DMRs fall into different chromatin states, where methylation is inversely correlated with active, but not repressive histone marks. However, methylation is not correlated with the expression of associated genes. High-resolution single nucleotide polymorphism (SNP) genotyping of the five donors revealed evidence for a role of cis-acting genetic variation in establishing methylation patterns. To validate this finding in a larger cohort, we performed genome-wide association studies (GWAS) using SNP genotypes and 450k array methylation data from blood samples of 1128 individuals. Only 30/157 (19%) DMRs include at least one 450k CpG, which shows that these arrays miss a large proportion of DNA methylation variation. In most cases, the GWAS peak overlapped the CpG position, and these regions are enriched for CREB group, NF-1, Sp100 and CTCF binding motifs. In two cases, there was tentative evidence for a trans-effect by KRAB zinc finger proteins.
Allele-specific DNA methylation occurs in discrete chromosomal regions and is driven by genetic variation in cis and trans, but in general has little effect on gene expression
A second poster from our working group will be presented at ECCB 2016. Daniela Beisser will present a poster about Taxonomic assignment of protist metatranscriptome sequences. She will also present the topic during the ECCB workshop “W11 – Recent Computational Advances in Metagenomics (RCAM’16)” on 4th September. See the workshop website for more information.
Taxonomic assignment of protist metatranscriptome sequences
Daniela Beisser, Nadine Graupner, Lars Grossmann, Jens Boenigk and Sven Rahmann
Next generation sequencing (NGS) technologies are increasingly applied to analyse complex microbial ecosystems by mRNA sequencing of whole communities, also known as metatranscriptome sequencing. In principle, each sequenced mRNA allows to both identify the species of origin and assign a function to the transcribed gene. While the functional information is sufficiently covered by databases such as Uniprot, NCBI, KEGG and many others, species identification is currently limited by incomplete reference databases. Inferring the community composition from metratranscriptomic samples is thus still a difficult problem. At the moment, most analyses are restricted to prokaryotic communities, which enjoy better database coverage, or to communities of few known species with sequenced genomes, or to a combination of rRNA and mRNA sequencing. However, the latter approach does not allow to link taxonomic and functional information directly.
Our approach focuses on an accurate assignment of taxonomic groups to metatranscriptomic reads. We constructed a custom database that comprises all major eukaryotic groups, developed a stand-alone tool to assign reads with a low false discovery rate and created a workflow for complete metatranscriptome analysis. The workflow covers all bioinformatic steps: preprocessing of the raw data, taxonomic and functional assignment, and visualisation of the results.
A poster about the Exome Analysis GraphicaL Environment (EAGLE) was accepted for the ECCB 2016 at The Hague. Felix Mölder will present the poster there.
EAGLE: an easy-to-use web-based exome analysis environment
Christopher Schröder, Felix Mölder, Christoph Stahl and Sven Rahmann
High throughput exome sequencing is a widely used technology for deciphering mutations in the coding regions of a genome at relatively low cost. While bioinformatics analyses of exome sequencing data mostly agree on best practices regarding the analysis steps, called genomic variants depend on the set of parameters and applied filtering. We present EAGLE, a software that combines a best practices variant calling workflow with a web frontend. By storing the called variant information in HDF5 files (instead of SQL databases), EAGLE allows filtering and parameter tuning in almost real time. This enables iterative tuning of thresholds, or the selection of different samples for filtering by medical PIs via the web interface. The web interface presents metadata, annotations, quality control data and statistics to facilitate a comprehensive data analysis on different levels.