Projects 2021

Projects for the Summer Internship Program 2021

This year we can offer 12 projects, which are as diverse as the research field Epigenetics, including topics such as:

histone modifications, environmental stress, nuclear architecture, image processing, metabolic diseases, cell fate decisions, deep learning, nucleosome positioning, totipotency, and spacial transcriptomics.

Even though, the situation of the COVID-19 pandemic is hard to predict for summer, we are optimistic that, with enough preparation time, the internship can be partially on campus. Thus, you have the chance to apply for experimental projects. Three projects will be on-campus only, four projects will be preferably on-campus and can be adapted to the virtual setting, and five projects are exclusively virtual.

The virtual setting was tested successfully in 2020 (see here). However, this year we will bring it to another level including more social activities and networking possibilities for all interns. One year of virtual meetings have shown us that this is possible.

In vivo study of chromatin organization upon stress in C. elegans (Cabianca Lab; IFE)

Multiple layers of regulation are required to establish and maintain appropriate gene expression patterns. These include chromatin modifications and higher order architecture of the genome. Epigenetic mechanisms can link the environment and the genome, yet how epigenomes respond to environmental perturbations remains largely unknown.

In our team, we study how nutrient stress affects the state, spatial compartmentation and function of chromatin within an intact developing organism, namely the roundworm C. elegans. Worms develop rapidly, are genetically manipulatable and are transparent, a feature that gives us the unique possibility to monitor fluorescently labeled chromatin within the cells of an intact animal.

In this internship, you will familiarize with C. elegans handling and genetics by performing a cross and image the resulting strain under a microscope. Also, you will contribute to the cloning of a tagged chromatin protein within an expression-suitable vector, thus gaining experience in molecular biology techniques. 

Supervisor: Dr. Daphne Cabiana

Institute: Institute for Functional Epigenetics

Setting: on-campus

Integration of single cell RNA-seq and ATAC-seq data (Colomé-Tatché Lab; ICB)

Recent technological advances allow for measuring the gene expression and chromatin openness of the same single cells, one cell at the time. These measurements open the possibility to study how epigenetic mechanisms affect gene expression at an unprecedented resolution.

However, mixing the information from the two molecular layers, epigenome and transcriptome, is not always trivial. Several methods and pipelines have been developed recently to address this question. In this project, you will work with a single cell RNA-seq and ATAC-seq dataset and will apply different methods for their joint analysis. Using a set of metrics you will evaluate the performance of the different integration methods.

Supervisor: Dr. Maria Colomé-Tatché

Institute: Institute for Computational Biology

Setting: on-campus/virtual

Single-locus specific chromatin isolation as a tool to study nucleosome positioning at the single-molecule level (Hamperl Lab; IES)

Chromatin presents the natural substrate of the essential and complex machineries dedicated to transcription, replication and repair. These fundamental processes require major chromatin rearrangements to access and make use of the genome. Thus, to understand the molecular basis of these DNA transactions, it is critical to define the collective changes of the chromatin structure at precise genomic regions where these machineries assemble and drive biological reactions.

This project will make use of an established affinity purification protocol to enrich specific chromosomal domains with high yields and purity (Hamperl et al, 2014), allowing us to purify a single-copy gene locus of interest in its native chromatin context. As part of this project, we will take advantage of this method and combine it with methylation-footprinting as well as other biochemical and functional assays to determine at the single-molecule level the heterogeneity and influence of nucleosome positioning and occupancy on the functional state of the locus of interest (transcription start sites and DNA replication origins).

Supervisor: Dr. Stephan Hamperl

Institute: Institute for Epigenetics and Stem Cells

Setting: on-campus

Redox-dependent chromatin modulation in environmental stress response (Lindermayr Lab; BIOP)

Redox molecules, such as nitric oxide and reactive oxygen species are signaling molecules with multiple regulatory functions in plant physiology and stress response. We are analyzing how plants respond to and cope with environmental stress conditions related to climate change. The response includes initiation of redox-signaling, epigenetic processes (e. g. histone modifications and DNA-methylation) and transcriptional reprogramming.

Using redox-mutants, we aim to identify redox-regulated histone modifications and genes regulated by such modifications (ChIP-seq). Moreover, the expression of these genes will be analysed (RNA-seq). Additionally, we want to identify redox-regulated chromatin modifiers, e. g. histone acetyltransferase/deacetylase and histone methyltransferases/demethylases and characterize their physiological function in environmental stress response.

Supervisor: Dr. Christian Lindermayr

Institute: Institute of Biochemical Plant Pathology

Setting: on-campus

Image processing and deep learning for live red blood cell fluorescence microscopy (Marr Lab; ICB)

Calcium is a universal signaling molecule playing an important role in regulating cell structural integrity, motility, volume, cell cycle and fate. Red blood cells (RBC) are dependent on Ca2+ during differentiation: “Malfunction of Ca2+ transporters in human RBCs leads to excessive accumulation of Ca2+ within the cells. This is associated with a number of pathological states including sickle cell disease, thalassemia, phosphofructokinase deficiency and other forms of hereditary anaemia.” [1] Calcium imaging is a microscopy method to measure the amount of calcium in cells.

Since the cells are not adhered to any surface and Ca2+ fluorescence images are obtained a few seconds after the brightfield image, due to movement of the cells registration of both images is desirable. Measurement of the amount of calcium in the cell and counting the number of vesicles, visible as small blobs of liquid, provides great insights to biologists about the RBC status, and eventually the state of the disease.  

In this project we want to develop image processing methods to register both channels, measure the intensity of the calcium fluorescence for every cell and with deep learning architectures detect if any vesicle is formed on the cell membranes and count them.

Milestones

  • Single cell detection in Ca2+ channel
  • Associating the detected cells in Ca2+ channel to those in brightfield
  • Measurement of the intensity in Ca2+ channel per cell
  • Detection of vesicles
  •  

Requirements

  • Basic python knowledge
  • Interest in microscopy image analysis

References

     

  1. Bogdanova, Anna, Asya Makhro, Jue Wang, Peter Lipp, and Lars Kaestner. "Calcium in red blood cells—a perilous balance." International journal of molecular sciences 14, no. 5 (2013): 9848-9872.

Supervisors: Ario Sadafi and Dr. Carsten Marr

Institute: Institute for Computational Biology

Setting: Virtual

Characterization of multi-species developing epigenetic landscapes (Marr Lab; ICB)

Embryonic development of mammals follows a strict pattern. However, the speed of these events, i.e. the developmental timescale, is distinct for each species and vastly differs within a taxonomic class [1]. Single cell sequencing offers the great potential to scrutinize cell compositions in an unprecedented resolution, providing insights into developmental processes that cannot be overserved in bulk sequencing approaches alone.

Our main goal is to uncover the molecular mechanisms that determine species-specific timescales of differentiation on the cellular level. We will employ longitudinal single-cell ATAC sequencing [2] of multiple mammalian species to identify key epigenetic differences between species during differentiation in order to understand the role of time in development.

In this project we want to develop image processing methods to register both channels, measure the intensity of the calcium fluorescence for every cell and with deep learning architectures detect if any vesicle is formed on the cell membranes and count them.

Milestones

  • Single cell ATAC data preprocessing (quality control, normalization, dimension reduction)
  •  

  • Cell annotation, matching ATAC and RNA signals
  •  

  • Comparison of epigenetic landscapes between species
  •  

  • Analysis of chromatin accessibility changes over time

Requirements

  • Interest in computational biology 
  •  

  • Basic knowledge of python and R

References

1. Ebisuya M, Briscoe J. (2018). What does time mean in development?. Development.

2. Baek S, Lee I. (2020). Single-cell ATAC sequencing analysis: From data preprocessing to hypothesis generation. Comput Struct Biotechnol J. 

Supervisors: Moritz Thomas and Dr. Carsten Marr

Institute: Institute for Computational Biology

Setting: Virtual

Uncertainty estimation from latent space representations using variational autoencoders and comparison to established methods (Marr Lab; ICB)

Estimating the uncertainty of neural network predictions is important, in particular in medical and clinical applications. Existing approaches such as MC dropout or augmentation based sampling require extra computation time, because the model needs to be evaluated numerous times [1, 2].

Variational autoencoders have the advantage of a regularized latent space, from which we want to estimate the uncertainty directly and compare this measure to existing methods. 

Milestones

  • Implement a variational autoencoder and add a latent space sampling from which uncertainty can be estimated compared to known uncertainty estimation metrics
  • Test our approach on different biomedical datasets 

Requirements

  • Interest in computational biology
  • Knowledge of python
  • Experience in deep learning algorithms

References

[1] Moloud Abdar et al., A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges, https://arxiv.org/pdf/2011.06225.pdf (2020)

[2] Xuming Ran et al.,Detecting Out-of-distribution Samples via Variational Auto-encoder with Reliable Uncertainty Estimation, https://arxiv.org/pdf/2007.08128.pdf (2020)

Supervisors: Dominik Waibel and Dr. Carsten Marr

Institute: Institute for Computational Biology

Setting: Virtual

Deep learning-based super-resolution Plankton imaging (Peng Lab; Helmholtz AI)

Plankton imaging systems, like all microscopic systems, are limited by the trade-off between magnification and depth-of-field (DOF) that restricts analysis to very small sample volumes. Recent technical  advances in opto-electronics, such as electrically tunable lenses (ETL), show great promise to overcome this challenge by substantially enhancing the DOF. However, acquired images still suffer from sub-optimal optical resolution due to the interference of in-focus and out-of-focus objects during image formation [1].

The proposed project will explore the potential of AI-based solutions for maximizing the optical resolution of enhanced-DOF images acquired with modern ETL technology, to achieve “super-resolution imaging” in unprecedentedly large sample volumes. Particularly, we would like to use recently-proposed ‘plugin-and-play’ approach for image deconvolution and super-resolution reconstruction [2]. 

Milestones

  • Estimate blur kernels from calibration images
  • Image deconvolution using traditional image deconvolution plugin of ImageJ [3]
  • Develop deep learning approaches based on [2]

Requirements

  • Strong mathematical/computational background
  • Experience in deep learning and Pytorch framework

References

[1] Taucher et al., “In situ camera observations reveal major role of zooplankton in modulating marine snow formation during an upwelling-induced plankton bloom,” Prog. Oceanogr., vol. 164, pp. 75–88, May 2018.

[2] Zhang et al. “Deep Plug-and-Play Super-Resolution for Arbitrary Blur Kernels” CVPR 2019. https://github.com/cszn/DPSR

[3] Sage et al. “DeconvolutionLab2: An Open-Source Software for Deconvolution Microscopy” Methods—Image Processing for Biologists, vol. 115, pp. 28-41, February 15, 2017. http://bigwww.epfl.ch/deconvolution/deconvolutionlab2/

Supervisors: Tingying Peng and Jan Taucher

Institutes: Helmholtz AI, Helmholtz Zentrum München and Geomar, Helmholtz Zentrum for Ocean Research Kiel

Setting: Virtual

Deciphering the function of novel histone modifications and their role in metabolic diseases – integrative omics analysis (Schneider Lab; IFE)

Our aim is to understand how novel types and sites of histone modifications regulate cellular functions and their deregulation in diseases such as cancer or diabetes (see for examples: Kebede et al., 2017; Tropberger et al., 2013; di Cerbo et al., 2014, Shahidian et al., 2021). By using a combination of different “omics” approaches and experimental manipulations our lab is identifying novel pathways regulating chromatin function in order to discover new therapy targets and unique diagnostic or prognostic markers.

The project for the summer student will focus on the analysis of “omics” data (ChIPseq, CUT&Tag, ATACseq and RNAseq) to unravel how histone modifications can regulate genome function and disease processes. This includes the possibility to be (i) part of a very dynamic and international team, (ii) learn state of the art data analysis technologies (including data from samples with very low cell numbers and integrated analysis of multi-omics data to leverage the full power of the data) and (iii) get insights in the fascinating field of the cross-talk between gene regulation and the environment.

Supervisor: Prof. Robert Schneider

Institute: Institute for Functional Epigenetics

Setting: on-campus/virtual

Physical models of cellular fate decision in olfactory sensory neurons (Scialdone Lab; IES/IFE/ICB)

In animals, the olfactory sensory system is fundamental for survival and reproduction. Most mammals are equipped with a complex olfactory system, where each of the cells responsible for odorant detection – the Olfactory Sensory Neurons (OSNs) – activates only one, randomly chosen olfactory receptor gene (OR) out of thousands, by means of a largely unknown mechanism.

The project consists in the analysis of spatial transcriptomic and chromatin 3D organization datasets from OSNs by using machine learning methods. Based on the results of these analyses, we will start building a physical model that can explain the OR gene random choice.

Supervisor: Dr. Antonio Scialdone

Institute:  Institute for Epigenetics and Stem Cells,  Institute for Functional Epigenetics and  Institute for Computational Biology

Setting: on-campus/virtual

Deep learning and dropout to select informative features for spatial transcriptomics (Theis Lab; ICB)

Single cell technologies have uncovered cellular heterogeneity across diseases, tissues, and organisms. Yet, only with the advent of spatial transcriptomics have we been able to localize processes described via other single-cell methodologies. The highest resolution spatial protocols are targeted and thus require a pre-selection of genes to use as probes for sequencing. Typically, known marker genes are chosen as probe sets. Yet, in order to capture variation beyond cell types, these genes can also be selected to be maximally informative of the variation in scRNA-seq data from the same tissue.

In this internship, we will use deep-learning-based methods for latent space learning to select a minimal set of features that describe the variation in a scRNA-seq dataset. Building on previous work for deep-learning-based feature selection, we will use an auto-encoder architecture to generate a latent space embedding and use dropout to select features. We will explore dropout and forward search algorithms to select features and compare these approaches to methods based on network gradients.

Supervisors: Dr. Malte Lückenl and Prof. Fabian Theis

Institute: Institute for Computational Biology

Setting: Virtual

Uncover the epigenetic mechanisms behind the establishment of totipotency (Torres-Padilla Lab; IES)

The project proposed aims at identifying the molecular players and epigenetic mechanisms that allow early embryonic cells to acquire their high plasticity or totipotency. Cellular plasticity is the capacity of a cell to give rise to different cell types upon differentiation.

In particular, the student will work with a cell culture model for totipotency termed 2-cell-like cells, so-called because they resemble the totipotent 2-cell stage mouse preimplantation embryo at the molecular level. They will implement biochemical and molecular biology approaches to understand the function of candidate proteins and RNAs in the maintenance and reprogramming to totipotency. 

The student will be able to grow and culture mouse embryonic stem cells and 2-cell-like cells, and manipulate the cells using RNAi, CrisprCas9-based epigenetic engineering and ectopic expression of chromatin modifiers. Overall, the project proposed aims to uncover the epigenetic mechanisms behind the establishment of totipotency and to provide insights into the origin of first pluripotent stem cells to form.

Supervisors: Dr. Adam Burton and Prof. Maria-Elena Torres-Padilla

Institute: Institute for Epigenetics and Stem Cells

Setting: on-campus/virtual