Logo Logo
Switch language to English
Zacher, Benedikt (2016): Genomic data integration with hidden Markov models to understand transcription regulation. Dissertation, LMU München: Fakultät für Chemie und Pharmazie



Transcription is a tightly controlled process that involves the recruitment and prost-translational modification of DNA-associated protein complexes, which can be mapped to the genome using high-throughput experimental assays. An accurate annotation of genomic elements such as transcription units or cis-regulatory elements such as promoters or enhancers is crucial for the use and interpretation of data generated by these assays. Thus, integrative genomic data analysis of high-throughput assays with hidden Markov models (HMMs) has become a popular tool for genome annotation. However, current algorithms are limited by unrealistic data distribution assumptions and variance models. Moreover, they are not able to assign forward or reverse direction to states or properly integrate strand-specific (e.g., RNA expression) with non-strand-specific (e.g., ChIP) data, which is essential to characterize directed processes such as transcription. In this thesis new HMM-based methods are proposed to overcome these limitations. These include (i) bidirectional HMMs (bdHMMs) which integrate strand-specific with non-strand-specific data to infer directed genomic states de novo and (ii) GenoSTAN (Genomic STate ANnotation), a HMM using discrete probability distributions to model count data, for genome annotation from Next-Generation-Sequencing data. Both approaches are made available in the R/Bioconductor package STAN (STate ANnotation) which provides an efficient implementation that can be run on large genomes such as human. STAN is used to derive new and improved annotations of transcription in yeast and human and to generate a map of promoters and enhancers in 127 human cell types and tissues.Integration of transcription factor binding and RNA expression data in yeast recovers the majority of transcribed loci, reveals gene-specific variations in the yeast transcription cycle, identifies 32 new transcribed loci, a regulated initiation-elongation transition, the absence of elongation factors Ctk1 and Paf1 from a class of genes, a distinct transcription mechanism for highly expressed genes and novel DNA sequence motifs associated with transcription termination.Moreover, promoters and enhancers are predicted in 127 human cell types and tissues are mapped by integrating sequencing data from the ENCODE and Roadmap Epigenomics projects, today’s largest compendium of chromatin assays. Promoters and enhancers are identified with consistently higher accuracy and show significantly higher enrichment of complex trait-associated genetic variants than current annotations. Investigation of binding of 101 transcription factors in human K562 cells reveals common and distinctive TF binding properties of enhancers and promoters.Application of STAN to transient transcriptome sequencing (TT-Seq) data in human K562 cells recovers stable mRNAs, long intergenic non-coding RNAs, and additionally maps over 10,000 transient RNAs, including enhancer RNAs, antisense RNAs, and promoter-associated RNAs. Further analyses reveal that transient RNAs such as enhancer RNAs are short and lack U1 motifs and secondary structure. Taken together, the annotations inferred in this thesis gave new insights into transcription and its regulation and will be an important resource for future research in genomics. STAN is a valuable tool to create such annotations also in other organisms and as more data becomes available improve the existing ones.