Life Science Data Analytics
and Algorithmic Bioinformatics


Our research is driven by the idea that modern data sciences can contribute significantly to address important questions in medicine and life sciences. Algorithms from the fields of statistical learning and data mining can be used to support the causal understanding, diagnosis and prognosis of complex diseases. Artificial intelligence methods can be applied to reverse engineer complex biological systems and to infer properties of these systems. This in turn could help to identify novel therapies in the future.

In the following selected ongoing and past research projects are outlined.

To top

Predicive Modeling of Health Claims Records (UCB)

Using big data (health claims records) analytics we investigate, whether future disease states and comorbidities of patients can be predicted on an individualized basis in order to optimize therapy. The project involves a larger effort to integrate and link different data sources in order to capture as best as possible comorbidity characteristics as well as features of applied medications.

Involved PhD student: Thomas Gerlach

DFG Graduate School 1873 (Sub-Project 8)

The M2 receptor is a pharmacologically relevant target. Within the DFG funded graduate school 1873 we focus on characterizing and predicting the behavior of the M2 receptor dependent signaling network. For this purpose we utilize and develop methods from mathematical modeling as well as statistical inference.

Involved PhD student: Benjamin Engelhardt


  • Benjamin EngelhardtHolger Fröhlich, Maik Kschischo,
    Learning (from) the errors of a systems biology model, Nature Scientific Reports, 6, 20772, 2016.

To top

Identification of Novel Therapeutic Approaches via Analysis of Residual Cells (BMBF, closed)

Together with our project partners from the University Medical Center Bonn we investigated different cell populations in Glioblastoma Multiforme (GBM) based on patient gene expression profiles. The aim was to characterize these differences and to investigate novel therapeutic options. A further goal of the project was to come up with predictive models based on transcriptome signatures.

Involved PhD students:

  • Ashar Ahmad
  • Joao Dinis (alumni)
  • Satya Samal Swarup (alumni)

To top

Neuroalliance Project D9 (BMBF, closed)

Together with partners from the University Medical Center Bonn and UCB Pharma we investigated the potential of miRNAs as diagnostic biomarkers for epilepsy. Animal models (using different experimental conditions) as well as primary patient samples were investigated.

Involved PhD candidate: Khalid Abnaof


  • Anita Kretschmann; Khalid Abnaof; Marijke van Rikxoort; Kirsten Ridder; Holger Fröhlich; Benedicte Danis; Rafal Kaminski; Patrik Foerch; Christian Elger; Frank Weinsberg; Alexander Pfeifer,
    Changes in serum miRNAs following generalized convulsive seizures in human mesial temporal lobe epilepsy, Biochemical and Biophysical Research Communications, 481(1-2): 13 - 18, 2016
  • Anita Kretschmann, Benedicte Danis, Lidija Andonovic, Khalid Abnaof, Marijke van Rikxoort, Franziska Siegel, Manuela Mazzuferi, Patrice Godard, Etienne Hanon, Holger Fröhlich, Rafal M. Kaminski, Patrik Foerch, Alexander Pfeifer,Different microRNA profiles in chronic epilepsy versus acute seizure mouse models, J Mol Neuroscience, 55(2):466-79, 2015 (awared with Neuroallianz Silver medal)
  • Anita Kretschmann; Khalid Abnaof; Marijke van Rikxoort; Kirsten Ridder; Holger Fröhlich; Benedicte Danis; Rafal Kaminski; Patrik Foerch; Christian Elger; Frank Weinsberg; Alexander Pfeifer,
    Changes in serum miRNAs following generalized convulsive seizures in human mesial temporal lobe epilepsy, Biochemical and Biophysical Research Communications, 481(1-2): 13 - 18, 2016

To top

Alma-in-Silico (EU, closed)

Together with researchers from the RWTH Aachen we investigated the effect of TGF-beta stimulation in several human and murine cell types based on gene expression time series data. We were able to identify commonly affected biological pathways and networks from these complex data.

Involved PhD student: Khalid Abnaof


  • G. Walenda, K. Abnaof, S. Joussen, S. Meurer, B. Smeerts, B. Rath, K. Hoffmann, H. Froehlich, R. Weiskirchen, M. Zenke, W. Wagner,
    TGF-beta1 Does not Induce Senescence of Mesenchymal Stromal Cells and has Similar Effects in Early and Late Passages, PLoS ONE, 8(10): e77656, 2013.
  • Khalid Abnaof, Nikhil Mallela, Gudrun Walenda, Steffen Meurer, Kristin Sere, Qiong Lin, Bert Smeets, Kurt Hoffmann, Wolfgang Wagner, Martin Zenke, Ralf Weiskirchen, Holger Froehlich,
    TGF-beta Stimulation in Human and Murine Cells Reveals Commonly Affected Biological Processes and Pathways at Transcription Level, BMC Systems Biology, 8:55, 2014.

To top

B-IT Graduate School (NRW, closed)

Learning Molecular Networks from Perturbation Data

This PhD project focused on new methods to learn molecular networks from downstream effects of targeted perturbations based on the framework of Nested Effects Models (NEMs). Particular achievements were:

  • An approach to integrate information from different, heterogenous data sources (e.g. GO, KEGG, InterPro, protein-protein interactions) as prior knowledge into the network inference procedure
  • A computationally efficient extension of static to Dynamic Nested Effects Models
  • Learning molecular networks from time lapse microscopy data

Involved PhD student: Paurush Praveen


  • H. Froehlich, P. Praveen, A. Tresch,
    Fast and Efficient Dynamic Nested Effects Models, Bioinformatics, 27, 238-244, 2011.
  • P. Praveen, H. Froehlich,
    Boosting Probabilistic Graphical Model Inference by Incorporating Prior Knowledge from Multiple Sources, PLoS ONE, 8 (6):e67410, 2013.
  • H. Failmezger, P. Praveen(*), A. Tresch, H. Froehlich,
    Learning Gene Network Structure from Time Laps Cell Imaging in RNAi Knock-Downs, Bioinformatics, 29(12):1534-40, 2013.(* joint first authorship).


Integrating Prior Knowledge Into Predictive Models for Biomarker Signature Discovery

The focus of this PhD project was on the question, how biological background knowledge (e.g. protein-protein interactions) could be integrated into predictive models for biomarker signature discovery in order to enhance the stability as well as interpretability of signatures. Particular achievements were:

  • An extensive review and comparison of existing methods using network information
  • Development of a novel approach integrating gene expression and miRNA expression as well as protein-protein interactions into one SVM classification model
  • Development of a software (R-package) integrating the new as well as several existing methods

Involved PhD student: Yupeng Cun


  • Y. Cun, H. Froehlich,
    Biomarker Gene Signature Discovery Integrating Network Knowledge, Biology, 1(1), 5-17, 2012 (Review)
  • Y. Cun, H. Froehlich,
    Prognostic Gene Signatures for Patient Stratification in Breast Cancer - Accuracy, Stability and Interpretability of Gene Selection Approaches Using Prior Knowledge on Protein-Protein Interactions, BMC Bioinformatics, 13:69, 2012.
  • Y. Cun, H. Froehlich,
    Network and Data Integration for Biomarker Signature Discovery via Network Smoothed T-Statistics, PLoS ONE, 8(9):e73074, 2013.
  • Y. Cun, H. Froehlich,
    netClass: An R-package for network based, integrative biomarker signature discovery, Bioinformatics, 30(9):1325 - 1326, 2014.

To top