ComputationalEpigenomics |

Prediction of single cell gene expression

Fri, 21 Aug 2026 00:00:00 +0000

Learning the relationship between gene-regulatory regions from epigenome and transcriptome data measured for single-cell is hard. We have developed an automated Nextflow pipeline that addresses the noise and sparsity of single-cell data by aggregating individual cells into metacells. We show that analysis on metacell level often leads to superior models and we illustrate the use of these models on different datasets.

Linked publications

Laura Rumpf, Fatemeh Behjati Ardakani, Dennis Hecker, Marcel H. Schulz (2026). Predicting gene-specific regulation with transcriptomic and epigenetic single-cell data. Bioinformatics.

DOI

Enabling allele-specific analysis of epigenome data

Fri, 19 Jun 2026 00:00:00 +0000

Project Description

Linking enhancers to genes is difficult. The state-of-the-art appraoch to link regulatory regions to genes is using the generalized Activity-by-Contact approach (gABC) see . The thesis is about enhancing the exisiting workflow to handle allele-specific epigenome data.

Research Goals

Adjust the to accommodate allele-specific epigenome signals
Test the new pipeline on novel heart related epigenome data

Prerequisites

Python programming
ideally some knowledge about epigenomics

How to Apply

Send a short Email with a short motivation to the corresponding supervisor .

Expanding a Nextflow pipeline for histone-association studies

Fri, 19 Jun 2026 00:00:00 +0000

Project Description

The epigenome is the layer on top of the genome that regulates how genes are expressed. We developed a Nextflow pipeline that uses machine learning models, that were trained to predict gene expression from epigenome data, to find genes which are associated with specific conditions such as cancer.

Research Goals

Adjust and expand the to find which regions differ between groups of samples.
Add a test to find which transcription factors are enriched in those regions.
Test the pipeline on heart related epigenome data.

Prerequisites

Python programming
Ideally some knowledge about epigenomics
Experience with Nextflow would help but not necessary

How to Apply

Send a short Email with a short motivation to the corresponding supervisor .

Expanding a Nextflow pipeline for single-cell data

Fri, 19 Jun 2026 00:00:00 +0000

Project Description

Research Goals

Adjust and expand the to also allow the usage of single-cell data.
Add input processing of single-cell data, which can then be given to pre-trained single-cell models.
Test the pipeline on additional data.

Prerequisites

Python/R programming
Ideally some knowledge about epigenomics
Experience with Nextflow would help but not necessary

How to Apply

Send a short Email with a short motivation to the corresponding supervisor .

MetaFR

Fri, 17 May 2024 00:00:00 +0000

TODO: add description and logo

Enhancer-gene interactions and transcription factor affinities

Thu, 26 Oct 2023 00:00:00 +0000

Enhancer–gene interactions are an important factor in epigenomics because they help to explain how regulatory regions control gene expression. Enhancers can act over large distances and do so in a cell type-specific manner. These interactions help connect epigenomic features to functional gene regulation, by finding regulatory programs that define cellular identity or provide interpretation for disease-associated noncoding variants.

We developed STARE, a computational tool for assigning enhancers to their likely target genes and using those to summarize transcription factor affinity on gene-level. The enhancer-gene interactions are predicted with the generalized Activity-by-Contact (gABC)-score, which uses a measurement of enhancer activity and optionally also chromatin contact data. It can be applied to pseudobulked single-cell data as well.

We applied STARE in a variety of different projects, especially the enhancer-gene interactions. Its applications so far include:

Examining the role of an understudied histone modification during cardiogenesis
Investigation of the relation between metabolism and epigenomics at the example of the cysteine pathway
Prioritizing disease genes for coronary artery disease by interpreting the impact of genomic risk variants on gene regulation
The role of STAT3 in pericytes under HFpEF

MISPIL and/or PCMultiome

Thu, 26 Oct 2023 00:00:00 +0000

We use gene-specific models to predict expression from epigenetic activity and perturb candidate regulatory elements in silico to identify small combinations that shift disease-associated expression profiles toward healthy states. By formulating this as a combinatorial optimization problem, we can efficiently prioritize regulatory regions for follow-up experimental perturbation.

Prediction of gene expression from epigenome data

Thu, 26 Oct 2023 00:00:00 +0000

Prediction of gene expression from epigenome data is an important task to link epigenome variation to changes in gene expression. Over the years we have designed a number of machine learning methods that present approaches for learning those links. We have made a number of models available for the community in the EpiExpress repository that allows to conduct histone-acetylation based epigenome association studies.

Linked publications

Fatemeh Behjati Ardakani, Shamim Ashrafiyan, Laura Rumpf, Dennis Hecker, Marcel H Schulz (2026). Comparing machine learning methods predicting transcriptome from epigenome with applications to association studies. Genome Biology.

URL DOI

Florian Schmidt, Alexander Marx, Nina Baumgarten, Marie Hebel, Martin Wegner, Manuel Kaulich, Matthias S. Leisegang, Ralf P. Brandes, Jonathan Göke, Jilles Vreeken, Marcel H. Schulz (2021). Integrative analysis of epigenetics data identifies gene-specific regulatory elements. Nucleic Acids Research.

DOI

Florian Schmidt, Fabian Kern, Marcel H. Schulz (2020). Integrative prediction of gene expression with chromatin accessibility and conformation data. Epigenetics & Chromatin.

DOI

STARE

Sat, 11 Sep 2021 00:00:00 +0000

STARE is a framework that combines two tasks, which can also work independently:

Score enhancer-gene interactions with the generalized Activity-By-Contact (gABC) model. It uses a measurement of enhancer activity and optionally also chromatin contact data.
Derive TF affinities for regions.

Those two parts are combined to summarise TF affinities on gene level, which allows for a variety of downstream analyses. It can be applied to pseudobulked single-cell data as well.