Question

How do developing tissues elongate?

Challenges

Morphogenesis is the biophysical process by which organisms develop their shapes. Understanding the mechanisms underlying morphogenesis requires precise analyses of cell dynamics within developing tissues. Tissue elongation along one axis is a simple and widespread process, and therefore ideal to start tackling these questions. In the absence of growth, it can be produced by the elongation of individual cells (either active, or in response to external forces), but also by oriented cell rearrangements. Such spatial rearrangements, by changing the topology of the system, allow net tissue deformation without accumulating cell deformation and stress. Delineating precisely the contributions of cell behaviours to the overall tissue elongation requires spatio-temporal quantification of cell shapes from experimental movies. The module will thus focus on the quantitative analysis of cell and tissue shape during tissue elongation in the Drosophila embryo.

Type of data

Movies of tissue elongation (Drosophila) from confocal microscopy

Methods

Image/movie segmentation, cell shape analysis, cellular force inference

Question

Can we detect non-coding transcription specific to cancers cell lines?

Challenges of the project

In our genome, genes are transcribed by the RNA-polymerase machinery, giving rise to transcripts that code for a proteins. The regulation of transcription is a vital process in all living organisms. Gene transcription is regulated by promoters and enhancers. Knowing how genes are regulated in a particular condition (eg: cancers, cell growth, etc…) has been a central challenge in biology and genomics. Recent findings have determined that non-coding regions, the non-genic part of our genome encompassing 80%, can also be transcribed by the RNA-polymerase machinery. This non-genic transcription may represent active regions involved the regulation of transcription.

In this project we will use the DNA binding of RNA-polymerase as a proxy for identifying “active” non-coding regions. Integrating this large amount of regulatory data across thousands of cells line will allow us to detect those cancer specific regulatory regions. We expect students to apply various mathematical and informatic analyses to first detect regions common across similar cancers, and second identify the most specific regions per cancer type.

Types of data

We have collected about 800 ChIP-seq dataset for RNA-Polymerase across a large variety of cell lines and cancer cell lines.  Foreach dataset, we have identified thousands of genomics regions bound by

RNA-Polymerase. The data consists of genomic regions (coordinates) that can be manipulated as a large matrix.

Methods

Matrix clustering, Dimensionality reduction, Gene Ontology enrichments, Intersection with genomic features (Genes, Retrotransposons, etc…).

Prerequisite

Student will apply dimension reduction using Python (R is also a possibility).

Question

How do B lymphocytes progress through multiple functional states to gain the ability to produce high affinity antibodies?

Challenges

Antibodies are lifesaving biomolecules – major effectors of vaccines efficacy and popular therapeutics for cancer treatment. In natural immunity, highly potent antibodies are produced by the descendants of a single antigen-specific B cell after a process of cellular and molecular maturation. In the aim of designing new therapies for challenging infections and cancers, understanding the distinct steps of B lymphocyte maturation is a major focus of immunology research.

Novel analytical methods enable the parallel analysis of antibody gene sequences and genome-wide gene expression in hundreds to thousands of single B cells, opening new avenues to dissect the mechanisms of B cell maturation. Integrating and analyzing the massive amount of data produced in those experiments will be an exciting challenge in the coming years.

Type of data

For thousands of B cells, sorted from healthy human lymphoid organs, human tumors, and mouse lymphoid organs after vaccination:

- Gene expression data (single-cell RNAseq)

- Cell surface phenotype and size/aspect characteristics (flow cytometry)

- Antibody gene sequence (computed from single-cell RNAseq)

Methods

Fluorescence-activated Cell Sorting (FACS), single-cell RNA sequencing, sequence alignment and mapping on genome, sequence reconstruction from short read RNAseq, dimensionality reduction and clustering, pseudotime analysis, cell – cell interaction networks, gene – gene interaction networks.

Question

How do collective neuronal dynamics relate to function?

Challenges

Most brain functions are supported by the activity of neuronal ensembles distributed across brain regions. Understanding the structure of these representations, how they are learned, and how they lead to behavior is one of the main challenges facing neuroscientists. In the recent years, technology, in particular imaging, high throughput electrophysiology and transcriptomics, have revolutionized our way to analyze structure and function in the brain and how it relates to behavior. Neuroscience is flooded by a deluge of complex data describing a system that is, itself, already complex. Never has the need been greater for theoretical models and data analysis. Future progress requires developing the ability to extract meaningful knowledge from these data through analyses and modeling.

Type of data

Activity dynamics from multiple neurons in awake behaving rodents (prefrontal cortex, hippocampus and striatum):

- Electrophysiological signals
- Calcium fluorescence movies

Methods

Two-photon and endoscopic calcium imaging, Image analysis, signal processing, functional connectivity, clustering

Question

When presented with images of a unique 3D object represented at different orientations, humans can easily determine the axes and angles of rotations linking the images. What are the neural mechanisms underlying this processing?

Challenges

We have recorded numerous neurons in visual areas of mice when the animals are presented with rotating objects. We would like to know whether the activity of some of the recorded neurons depends only on the orientation of the displayed object, and not on the details of the object shape, color, or texture. The set of all the possible orientations of 3D objects forms a group which is denoted SO(3). In mathematical terms, we would like to understand whether a set of neurons encodes solely a position on the manifold SO(3).

Type of data

We have the calcium imaging dataset of about 8000 neurons in the cortex comprising visual areas when an animal is presented with several Tetris-like 3D objects. We have several repeats of the same object exploring different trajectories on SO(3).

Methods

We have established a theory for the neural representation of 3D rotations. We will, therefore, try to make a connection between this theory and the dataset at hand. To do that, we will employ methods borrowed from machine learning, optimization, and group theory.