Quantitatively understanding the stochastic dynamics of gene expression requires measurements at the level of single cells. A common approach to follow the expression of genes in single cells and in real time is to make use of fluorescent reporter proteins and to record the cells' fluorescence by microscopy. However, this provides only an indirect readout of the biological processes that are of interest such as the regulation mechanisms at the promoter. A possible way to uncover the unobservable biological processes is to infer the hidden dynamics from the available data through the use of mechanistic models of gene expression. The goal of this project is to develop methods for state estimation and parameter inference for such models and to test these methods on real data.

In early development, regulation of transcription results in precisely positioned and highly reproducible expression patterns that specify cellular identities. How transcription, a fundamentally noisy molecular process, is regulated to achieve reliable embryonic patterning remains unclear. In particular, it is unknown how gene-specific regulation mechanisms affect kinetic rates of transcription, and whether there are common, global features that govern these rates across a genetic network. Quantitative measurements of nascent transcriptional activity in both living and fixed tissues are key in order to understand the underlying transcription kinetics and to make progress with these fundamental questions. The current project aims at constructing realistic minimalist models of transcription for different experimental and developmental contexts, using spatiotemporal gene expression activity data obtained from microscopic imaging of live biological tissues.

Because of the fast-growing number of assembled genomes available in the public repositories (GenBank/RefSeq), it is now quite common to end up with many genome assemblies that belong to the same species. In this current context, determining the type strain/species of any new species/genus requires to identify the representive one(s) within a large collection of genomes. The medoid of a genome collection being the one(s) whose sum of distances to all other genomes is minimal, it is an excellent candidate to be the representative genome of a collection. We therefore aim at procuring a bioinformatic tool able to quickly and accurately identify the medoid of any genome collections.

Cytokinesis is the last step of cell division and leads to the physical separation of the daughter cells. The purpose of this project is to infer the timing of abscission in various cell types by using few microscopic images of live cells and fluorescent markers.