Biostatistics
Projects
This is a selection of current and previous research projects. Code and more information is provided via the GitHub repository.
Prospective COVID-19 Cohort Munich (KoCo19)
Main contacts: Turid Frahnow, Mercè Garí, Ronan Le Gleut, Houda Yaqine, Christiane Fuchs
Under the lead of the Division of Infectious Diseases and Tropical Medicine, we participate in the prospective COVID-19 Cohort Munich (KoCo19) study. We contribute through statistical analyses and stochastic modelling, especially using stochastic differential equation models which account for disease transmission across regions and within households.
More information: Project webpage
References:
- Radon et al. (2020): Protocol of a population-based prospective COVID-19 cohort study Munich, Germany (KoCo19), BMC Public Health 20:1036
- Pritsch et al. (2021): Prevalence and Risk Factors of Infection in the Representative COVID-19 Cohort Munich, http://dx.doi.org/10.2139/ssrn.3745128
- Olbrich et al. (2021): A Serology Strategy for Epidemiological Studies Based on the Comparison of the Performance of Seven Different Test Systems - The Representative COVID-19 Cohort Munich, doi.org/10.1101/2021.01.13.21249735
Uncertainty Quantification - From Data to Reliable Knowledge
Main contacts: Lisa Amrhein, Houda Yaqine, Christiane Fuchs
How will the climate develop, how secure is our energy supply, and what chances does molecular medicine offer? The rapidly increasing amount of data offers radically new opportunities to address today’s most pressing questions of society, science, and economy: Data, outcomes and predictions are, however, subject to uncertainties. The goal of the project Uncertainty Quantification is to understand these uncertainties through methods of probability theory, and to include them into research and outreach. The project connects applied researchers from the four research fields Earth & Environment, Energy, Health, and Information among each other and with Helmholtz data science experts, as well as external university partners from mathematics and econometrics.
More information: Project webpage
Modeling and Bayesian Inference for Diffusions
Main contacts: Susanne Pieschner, Houda Yaqine, Christiane Fuchs
We are intimately involved in diffusion modelling and the development of Bayesian estimation techniques for diffusions. The application of diffusion processes to fluorescence microscopy data and single-cell data yields promising results and shows the potential of this approach.
References
- C. Fuchs (2013). Inference for Diffusion Processes with Applications in Life Sciences. Springer, Heidelberg.
- S. Pieschner, C. Fuchs (2020): Bayesian inference for diffusion processes: using higher-order approximations for transition densities, R. Soc. Open Sci. 7:200270
Spatial organization and scientific collaboration at the Helmholtz Zentrum München and the Bielefeld University
Main contacts: Hannah Marchi, Christiane Fuchs
It seems evident that spatial proximity between researchers may lead to more frequent or more intense collaboration than between scientists who work at large distance from each other. We hypothesize that the spatial organization within a research campus or even within a building influences interdisciplinary work. In a collaboration network study, we investigate which distance matters, how much researchers are influenced by people working around them and how scientific publishing changes depending on the heterogeneity among authors.
Outcomes of the study could provide valuable information about which spatial organization can foster (interdisciplinary) research and could be used for future plans of building structures.
We plan to extend the analysis to other research centers and universities in Munich like the Ludwig-Maximilians University (LMU) which has institutes spread all over the city. In this way it would be possible to compare different campus structures and their influence on scientific collaboration.
References:
H. Busen, C. Fuchs (2020): Modelling the impact of spatial proximity on scientific collaboration networks. Proceedings of the 35th International Workshop on Statistical Modelling (IWSM).
Estimation of Single-cell Heterogeneities from Cell Populations
- Mixture of cells in different regulatory states (left) and single-cell mixture density (right) (source: Christiane Fuchs, HMGU).
Main contacts: Lisa Amrhein, Mercé Garì, Christiane Fuchs
Even when appearing perfectly homogeneous on a morphological basis, tissues can be substantially heterogeneous in single-cell molecular expression. As such heterogeneities might govern the regulation of cell fate, one is interested in quantifying the heterogeneities in a given tissue.
Gene expression measurements of single cells would be most suitable to detect and further parameterize a heterogeneous population if the dataset was large and error-free. Unfortunately, such measurements are often expensive and subject to substantial technical noise. Instead of considering single-cell data, we randomly select small numbers of cells and measure the subpopulation average expression levels.
We investigate how heterogeneities can be detected from such data by application of statistical methods, and how the proportions, mean values and standard deviations of the groups of differently expressed cells can be estimated.
Application to measurements from human breast epithelial cells reveals the functional relevance of the heterogeneous expression of a particular gene.
Source code, an R package and a webtool are provided on the StochasticProfiling project website.
Collaboration partner:
Prof. Dr. Kevin Janes, University of Virginia
References
- S. Bajikar, C. Fuchs, A. Roller, F. Theis and K. Janes (2014): Parameterizing cell-to-cell regulatory heterogeneities via stochastic transcriptional profiles. PNAS 111(5), E626-635.
- C. Fuchs (2014): Mische und herrsche. Laborpraxis 38, 20-22.
- C. Kurz (2015): Stochastic profiling in single-cell heterogeneities. A Bayesian Extension. MSc thesis, University of Applied Sciences Munich.
- S. Tirier, J. Park, F. Preußer, L. Amrhein, Z. Gu, S. Steiger, J.-P. Mallm, M. Waschow, B. Eismann, M. Gut, I.G. Gut, K. Rippe, M. Schlesner, F. Theis, C. Fuchs, C. Ball, H. Glimm, R. Eils, C. Conrad (2019): Pheno-seq – linking visual features and gene expression in 3D cell culture systems. Sci Rep 9, 12367.
- L. Amrhein, C. Fuchs (2020): stochprofML: Stochastic Profiling Using Maximum Likelihood Estimation. R.arXiv2004.08809 (accepted by BMC Bioinformatics).
- L. Amrhein, C. Fuchs (2020): Stochastic profiling of mRNA counts using HMC. Proceedings of the 35th International Workshop on Statistical Modelling (IWSM).
Past Projects
Detecting Genetic and Environmental Risks for Childhood Asthma
Main contacts: Norbert Krautenbacher, Christiane Fuchs
Childhood asthma is a widespread disease. Many studies revealed that its onset is influenced by genetic and environmental factors like certain single nucleotide polymorphism (SNP) variants, family history or farming environment.
Our objective is to develop an asthma risk score especially for children between one and three years with which one can assess a child’s personal risk to develop the disease. The score should be based on few SNPs and the environmental variables. This shall allow a cost-efficient targeted treatment for exposed children.
Statistical aspects of this project are regularization and variable selection, gene-environment interactions, big data, inclusion of prior knowledge, stratification of the data, missing values, SNP imputation and validation.
Collaboration partners:
Prof. Dr. Erika von Mutius, Dr. Markus Ege, Prof. Dr. Bianca Schaub
Dr. von Hauner Children‘s Hospital
References:
- N. Krautenbacher, M. Kabesch, E. Horak, C. Braun‐Fahrländer, J. Genuneit, A. Boznanski, E. von Mutius, F. Theis, C. Fuchs, M. J. Ege (2020): Asthma in farm children is more determined by genetic polymorphisms and in non‐farm children by environmental factors, Pediatric Allergy and Immunology, 10.1111/pai.13436
- PAI Interview N Krautenbacher & B Moya (AACI, European Academy of Allergy and Clinical Immunology)
- ROC curves for models taking into account genetics (left), environment (middle), or both (right) (source: Norbert Krautenbacher, HMGU).
Risk Prediction for Prostate Cancer Patients
Main contacts at ICB: Ivan Kondofersky, Norbert Krautenbacher, Hagen Scherb, Christiane Fuchs
The Prostate Cancer DREAM Challenge attempted to improve survival prediction of prostate cancer patients. Participants were asked to build risk scores from a bulk of snapshot and longitudinal data tables within four months.
As "A Bavarian Dream" we participated in this challenge and finished up among the winning teams in both Subchallenges 1 and 2. Our work involved data and result management, data cleaning and preprocessing in close collaboration with a clinician, and model building ranging from classical Cox regression to machine learning, ensemble methods and model averaging. Final predictions were evaluated on an independent test set that had been withheld by the challenge organizers.
References
- F. Seyednasrollah, D. Koestler, T. Wang, S. Piccolo, R. Vega, R. Greiner, C. Fuchs, E. Gofer, L. Kumar, R. Wolfinger, K. Kanigel Winner, C. Bare, E. Neto, T. Yu, L. Shen, K. Abdallah, Kald, T. Norman, G. Stolovitzky, PCC-DREAM Community, H. Soule, C. Sweeney, C. Ryan, H. Scher, O. Sartor, L. Elo, F. Zhou, J. Guinney, J. Costello, Prostate Cancer DREAM Challenge Community (2017): A DREAM Challenge to Build Prediction Models for Short-Term Discontinuation of Docetaxel in Metastatic Castration-Resistant Prostate Cancer. Journal of Clinical Oncology Clinical Cancer Informatics: 1, 1-15.
- J. Guinney, T. Wang, T. Laajala, K. Winner, C. Bare, E. Neto, S. Khan, G. Peddinti, A. Airola, T. Pahikkala, T. Mirtti, T. Yu, B. Bot, L. Shen, K. Abdallah, T. Norman, S. Friend, G. Stolovitzky, H. Soule, C. Sweeney, C. Ryan, H. Scher, O. Sartor, Y. Xie, T. Aittokallio, F. Zhou, J. Costello and the Prostate Cancer Challenge DREAM Community (2017): Prediction of overall survival for patients with metastatic castration-resistant prostate cancer: development of a prognostic model through a crowdsourced challenge with open clinical trial data. The Lancet Oncology 18, 132-142. DOI: 10.1016/S1470-2045(16)30560-5.
- I. Kondofersky, M. Laimighofer, C. Kurz, N. Krautenbacher, J. Söllner, P. Dargatz, H. Scherb, D. Ankerst, C. Fuchs (2016): Three general concepts to improve risk prediction: good data, wisdom of the crowd, recalibration [version 1; referees: awaiting peer review]. F1000Research 5(2671). DOI: 10.12688/f1000research.8680.1.
More information:
- Documentation of our methods
- ICB news
- HMGU news (access for HMGU members only)
- TUM Maths news
- Challenge organizers' press release
- From left to right: Donna Pauler Ankerst, Norbert Krautenbacher, Michael Laimighofer, Christoph Kurz, Christiane Fuchs, Hagen Scherb, Ivan Kondofersky, Julia Söllner; not pictured: Philip Dargatz (source: HMGU).
Learning Classifiers on Biased Samples
- Selection process leading to biased sample (source: Christiane Fuchs, Eleni Tsalma, HMGU).
Main contacts: Norbert Krautenbacher, Christiane Fuchs
In many epidemiological applications, particular interest lies on the investigation of rare combinations of an exposure and a target variable. Representative samples from a population hence may not contain sufficiently many cases for a reliable analysis. For that reason, stratified samples are taken from the population to enrich the rare combinations. Well-known examples are case-control studies or two-phase studies. The enrichment comes at the cost of biased samples distorting estimates. We address issues arising in prediction on such biased samples, both for training and evaluation of a statistical model.
References
- N. Krautenbacher, F. Theis, C. Fuchs (2017): Correcting classifiers for sample selection bias in two-phase case-control studies. Computational and Mathematical Methods in Medicine, vol. 2017, Article ID 7847531, doi:10.1155/2017/7847531.
N. Krautenbacher, K. Strauß, M. Mandl, C. Fuchs (2018): R Paket sambia, available on cran.r-project.org
Computational Models of Neoplasmic Heterogeneities and Lineage Choice
Main contacts at Biostatistics: Lisa Amrhein, Christiane Fuchs
Acute myeloid leukemia (AML) often results from the myelodysplastic syndrome (MDS). Here, the differentiation hierarchy from hematopoietic stem cells to mature, functional cells is disturbed. AML patients, again, frequently carry a mixture of different cancer cell types, so-called subclones. This is reflected by a mixture of genomic signatures and heterogeneous transcriptome profiles. Applying statistical and dynamical models to data from our clinical and biological collaborators, we want to identify altered differentiation hierarchies of MDS subclones and characterize the development of related heterogeneities in AML while the tumor undergoes evolution.
This project is funded as Subproject A17 of the Collaborative Research Centre (CRC) 1243 "Genetic and Epigenetic Evolution of Hematopoietic Neoplasms".
Further reading:
- Webpage of CRC 1243:
http://www.sfb1243.biologie.uni-muenchen.de/index.html
- Article by Research Features
- Objectives of this project. Left: Inferring transcriptional and genomic heterogeneity in AML, that is the co-existence of at least two distinct cell populations within a tumor, and how it evolves over time. Right: Estimating differentiation and self-renewal rates in healthy and clonal differentiation hierarchies from healthy and MDS patient data (source: Christiane Fuchs, Carsten Marr, HMGU).