Research Projects

This section introduces current projects of the Biomedical Data Science group.

Gender differences in neurodegenerative diseases (GenderND project)

In both PD and AD, gender differences have been observed in the incidence and phenotypic manifestations of the disorder. Although these differences may result from diverse behaviors and lifestyles, previous studies suggest that the underlying causes are more complex and disease-specific, and involve hormonal and genetic influences. In the GenderND project, we investigate whether specific genetic factors contribute to the observed gender differences in neurodegenerative disorders. For AD, we are using animal models and molecular data from human biospecimens to study a candidate gene derived from the statistical analysis of multiple large-scale omics datasets from AD case/control studies. The corresponding sex-linked gene, ubiquitin-specific peptidase 9 (USP9), has a gender-biased activity in the human brain and encodes an enzyme reported to regulate the phosphorylation of MAPT, a protein thought to play a central role in AD pathogenesis. We had previously shown that the knockdown of USP9 results in a decreased MAPT gene expression in zebrafish embryos and in a human cell culture model. 

More recently, preliminary experiments in a mouse model for AD based on the injection of amyloid-beta oligomers into brain hippocampi revealed significant, gender-specific alteration patterns in USP9 and MAPT. Moreover, in a second AD model, brains of mice expressing different human APOE isoforms (APP/E3 and APP/E4 mice) also displayed gender-dimorphic differences in USP9 and MAPT gene expression. 

For PD, an initial analysis of gender-linked genes using an integrated analysis of omics datasets from PD case/control studies revealed multiple candidate disease genes with gender-specific changes in PD. These candidates are currently further investigated using analyses of PD-related changes in the surrounding molecular interaction network. The resulting information on disease-linked gender differences in the brain transcriptome can provide new insights to support the development of patient-tailored diagnostic and therapeutic approaches for the studied neurodegenerative disorders.

Mitochondrial endophenotypes of Parkinson’s disease (MitoPD project)

The goal of this multi-centre international project, lead by Prof. Thomas Gasser from the University of Tübingen in Germany, is to integrate genomic, transcriptomic and proteomic data to computationally stratify PD patients in terms of disease-related pathway and network alterations. We are particularly interested in a putative subgroup of patients with pronounced mitochondrial dysfunction, suggested by prior evidence from genetic and biochemical studies. We aim to specify this subgroup more precisely by robustly discriminating a postulated “mitochondrial endophenotype” of PD from other cases, using both omics data derived from large, carefully phenotyped patient cohorts as well as corresponding animal and cellular models, within an integrated computational and experimental approach. A first-stage validation will be performed by testing mitochondrial function in patient biomaterials in sub-cohorts of patients with predicted mitochondrial phenotypes. Results of this validation step will be used to improve the machine learning models. The model validity will be confirmed in further cohorts of familial and sporadic patients and genetically-defined at-risk individuals as well as in animal and cellular models of PD. Pathway-specific and MR spectroscopy-based biomarkers will then be developed, and stratified sub-cohorts will be identified for proof-of-concept clinical trials. Within the project consortium, the Biomedical Data Science group is responsible for the pathway- and network-based machine learning analysis of the omics data. We take part in the proposal and design of new validation experiments, and iteratively evaluate and improve our models in collaboration with the experimental groups.

Multi-dimensional stratification of Parkinson’s disease patients for personalized interventions (PDStrat project)

Current therapeutic approaches for PD help alleviate some of the major symptoms, but can neither halt nor slow the progression of the disease. As part of a multicenter collaborative research project, led by Prof. Peter Heutink from the German Centre for Neurodegenerative Diseases (DZNE), we are investigating new approaches for personalized interventions against PD, hypothesizing that previous clinical trials have failed for three main reasons: First, selection of potential drug targets is rarely based on robust biological evidence. Second, the disease process begins decades before clinical symptoms are observed and clinical trials on patients are therefore likely too late to reverse the neurodegeneration. Third, participants have been selected largely ignoring their underlying disease biology, phenotypic variation and their genetic risk profile, resulting in a very heterogeneous population of cases with very different progression of disease. Therefore, we aim to develop a novel concept for disease-mechanism based disease onset and progression prediction, and subsequent target discovery and patient stratification for Parkinson’s disease. Our goals are to identify novel targets based on biological evidence with matching precision cohorts for clinical trials that will allow for personalized therapeutic interventions based on genetic and genomic risk profiles and clinical subtypes.To reach our goal, we are using a systems biology approach by generating multi-dimensional clinical, genetic/genomic and biological risk and progression profiles for patients and at risk individuals and integrating these data in network models for target discovery and prediction and progression models for patient stratification. Within the project consortium, the Biomedical Data Science group is responsible for the machine learning and network analyses of omics data and clinical records. The project is funded by the Luxembourg National Research Fund (FNR) as part of the multilateral ERACoSysMed JTC2 2017 call.

Comparing alterations in neuroprotective and neurotrophic genes in age-related diseases (NeuroProDB project)

Proteins with neurotrophic and neuroprotective functions are thought to play a key role in age-related disorders, such as Alzheimer’s and Parkinson’s disease. Sequence variations in corresponding genes can affect the susceptibility for neurodegenerative disorders, and the drug-based activation of neuroprotective proteins is widely studied as a possible new treatment strategy. However, previously no central database had been available that captures the current information on neurotrophic/protective proteins in the literature and allows researchers to study the available experimental evidence for a protein’s trophic/protective actions. By using literature mining in combination with subsequent manual curation, we have recently created a web-based database of known neuroprotective/-trophic proteins ( The web-service provides details on the reported in vitro and in vivo evidence for neurotrophic and/or –protective functions in the peer-reviewed literature, and the condition-specificity of these functions. This information is complemented by details on relevant gene expression changes in common neurodegenerative disorders, on protein sub-cellular localizations, tissue-specific gene/ protein expression and protein-protein interactions. Currently, we are using the assembled database to investigate to which extent neuroprotective/trophic functions of a protein can be predicted by machine learning algorithms, and to study disease-associated network perturbations around neuroprotective genes in Parkinson’s and Alzheimer’s disease.

PERsonalised MedicIne Trials (PERMIT)

The PERMIT project, funded by H2020, aims to develop and disseminate recommendations for robust and reproducible personalised medicine research. Over the span of two years, PERMIT is performing a detailed mapping of current methodologies applied at every stage of the personalised medicine research pipeline and identifying gaps and key areas for the definition of standards.

A series of recommendations will be published, to help guide stakeholders in their assessment, funding, publication and support of personalised medicine research. The objective is to build consensus on standards that will respond to regulatory expectations, will produce high quality, reproducible and reliable results, and support the conduct of multinational clinical trials across Europe.

Machine learning and stratification
The increasing use of high-throughput data generation technology is promoting patient-centred personalised medicine. Personalised medical practice requires research on data-driven patient stratification. Biomarker stratification signatures may be used to define a new disease taxonomy, to refine diagnostic procedures or to propose more targeted treatments for each patient cluster.

The Work Package 4 (WP4) led by Assistant Prof. Enrico Glaab (University of Luxembourg / ELIXIR-LU) aims to develop guidelines to ensure biomedical relevance, robustness, reproducibility, and validity of algorithm-driven patient stratification. WP4 will provide an inventory of artificial intelligence methods for omics-based stratification and validation, assess the various machine learning approaches capable of identifying distinctive patient clusters and examine how common workflows could be optimised.

Contact: Enrico Glaab