RNA dynamics and biomolecular systems
Group leader : Hervé Isambert
Our research concerns the reconstruction, analysis and evolution of bio-networks at different scales and their implication on the organisms’ susceptibility to genetic diseases such as cancer. We develop quantitative methods and computational tools to infer and analyze causal graphical models from biological data.
1- Robust reconstruction of causal networks
We have recently developped a novel information-theoretic approach combining constraint-based and Bayesian frameworks to reliably reconstruct causal networks, despite inherent sampling noise in finite datasets (Affeldt et al UAI 2015; Affeldt et al BMC Bioinformatics 2015). This "3off2" approach is shown to outperform both constraint-based and Bayesian inference methods on a range of benchmark networks (Figure 1A). Applied to the reconstruction of the hematopoiesis regulation network based on recent single cell expression data, 3off2 is found to retrieve more experimentally ascertained transcriptional regulations than with other available methods (Figure 1B).
2- Ancient genome duplications and genetic diseases in vertebrates
In collaboration with the Camonis lab, we analysed the evolutionary constraints on the Ras-Ral signaling networks implicated in cancer (Figure 2). We investigated the evidence that the emerging properties of these signaling pathways might actually reflect their susceptibility to oncogenic mutations and thus their implication in cancer.
We found, in particular, that “dangerous” gene families, defined as prone to dominant deleterious mutations, have been greatly expanded through two rounds of whole genome duplication (WGD) in early vertebrates, Figure 3 (Affeldt et al Cell Rep 2012; Singh et al PLoS Comput Biol 2014; Malaguti et al Theor Popul Biol 2014). These findings highlight the importance of WGD-induced non-adaptive selection for the emergence of vertebrate complexity, while rationalizing, from an evolutionary perspective, the expansion of gene families frequently implicated in genetic disorders and cancers.
Hence, identifying gene duplicates retained from WGD, coined "ohnologs" after Susume Ohno, is central to better understand the evolution of vertebrates and their susceptibility to genetic diseases. We recently reported (Singh et al PLoS Comput Biol 2015), the identification of vertebrate ohnologs based on the quantitative assessment and integration of synteny conservation between six amniote vertebrates and six invertebrate outgroups. Such a synteny comparison across multiple genomes is shown to enhance the statistical power of ohnolog identification in vertebrates by overcoming lineage specific genome rearrangements. Ohnolog gene families can be browsed and downloaded at http://ohnologs.curie.fr/.
3- Evolution of large biomolecular networks
We are interested in the properties of large biomolecular networks and their evolution due to single gene and whole genome duplications, that occurred repeatedly in the course of eukaryote evolution, Figure 3.
Our early theoretical analyses focussed on duplication-divergence models to account for the generic properties of biomolecular networks, Figure 4. We have shown that duplication-divergence processes bring not only genetic novelty but also evolutionary constraints that restrict by construction the emerging properties of biomolecular networks. In particular, we demonstrated that networks with evolutionary conserved genes display also necessary topological properties by construction (such as hubs and scale-free degree distribution), Evlampiev et al PNAS 2008; Stein et al PRE 2011; Evlampiev et al BMC Syst Biol 2007. We are also interested in the evolution of transcription networks and study the regulatory conflicts that arise through duplication of transcription factors and autoregulators, Cosentino-Lagomarsino et al PNAS 2007.
All in all, it appears that evolutionary constraints, inherent to duplication-divergence processes, have largely controlled the overall topology and scale-dependent conservation of biomolecular networks.
4- Synthetic RNA regulatory networks and self-assembly of bacterial RNA.
We have developed and maintained a server for advanced RNA dynamic simulation (http://kinefold.curie.fr >100,000 online simulations; Xayaphoummine et al NAR 2005; PNAS 2003; Isambert et al PNAS 2000) and studied the properties of small regulatory circuits primary based on RNAs and their interactions (Dawid et al Phys Biol 2009; Xayaphoummine et al NAR 2007).
In particular, we have used synthetic biology approaches coupled to advanced RNA dynamics simulations (Kinefold server, movies 1 & 2) to design efficient RNA-based repressor (Figure 5A) and activator modules (Figure 5B). These modules control RNA transcription "on the fly" through simple RNA-RNA antisense interactions, Figure 5.
We also discovered that a small bacterial RNA of Escherichia coli could self-assemble, like many proteins do, to form long filaments as well as novel RNA-based nanostructures, Figure 6. This finding further extends the already great versality of natural RNA functions.
Two-year postdoc position available for a computer scientist, bioinformatician or physicist to reconstruct network from large scale genomic data.
Send CV and the names of two references to herve.isambert[at]curie.fr