Bordeaux Computational Biology and Bioinformatics
November 25-26, 2014
- 9:00-9:45 : Keynote speaker (Data integration)
Speaker : Lodewyk Wessels (NKI, Amsterdam)
Title : Molecular networks as determinants of response and outcome
The exact mechanisms involved in tumor development and therapy response are still largely unclear. Here we report on two computational approaches we developed to systematically unravel these genetic interactions based on high throughput datasets.
In the first approach, we developed a kernel-based, scale space approach to detect molecular interactions (co-occurrences and mutually exclusivities) from somatic variant and copy number data. We show that the number of aberrations per gene and per sample have a major influence on the results and propose a null-model to correct for these effects. We demonstrate the approach on the TCGA and METABRIC breast cancer cohorts. We identify no co-occurring aberrations but multiple mutually exclusive interactions, that represent a global breast cancer interaction network, shedding new light on breast cancer development and subtyping. We also performed a pan-cancer interaction analysis which reveals cancer type independent interactions.
Second, we propose a novel computational approach based on integer programming that infers logic combinations of discrete genetic events that predict the observed phenotype. The use of a logic formalism enables the formulation of intelligible models, facilitating speedy hypotheses generation. Here, we report on the application of this approach to the Wellcome Trust Sanger Institute 1000 cancer cell line panel. On this panel, genome wide mutation status, copy number, gene expression and methylation profiles have been recorded. The response to 400 anti-cancer therapeutics have also been recorded. Our models show that for most drugs, combinations of mutations explain the drug response better than single mutations. For example, of the eight BRAF inhibitors in the panel, the drug sensitivity to three of them is better explained using a logic combination of a BRAF mutation and one or more mutations in other genes. These results immediately suggest putative drug combination therapies.
- 10:05-10:20 : Local invited speaker (Network analysis)
Speaker : Sophie Colombié
Title : Modelling fluxes and metabolic network analysis of developing tomato fruits
Our aim is to better understand central metabolism and the way it influences accumulation and quality of plant biomass, through a systems biology approach centered on fruit development. Flux estimations calculated with models of metabolic networks represent a valuable alternative to flux measurements in complex systems such as fruits. In 2010 we performed a series of phenotyping experiments, in which and a large set of climatic and ecophysiological data and samples have been collected to analyse enzyme activities, the metabolome and structural traits. We have developed and validated two models describing fruit development.
The first model is a small-scale kinetic model describing realistic rate equations for each step of the metabolic network responsible for the interconversion and storage of sugars. It uses experimentally determined subcellular volumes and enzyme capacities as parameters. Fitting of sugar contents enabled the description of sucrose uptake and partitioning throughout fruit development, and most interestingly pinpointed the strong control exerted by vacuolar carriers on sugar storage. The model has been validated with independent data from literature and has been used to study metabolism in depth. One of the major findings evidenced by the model was the high accumulation of soluble sugars within the vacuole together with organic acids, enabling the osmotic-driven vacuole expansion that occurs during cell division (Beauvoit et al., Plant Cell 2014).
The second model, a medium-scale stoichiometric model including the balance of cofactors and energy allows calculating fluxes in central metabolism throughout fruit development. Concentrations of the main biomass components were determined at each developmental stage and fitted in order to calculate the corresponding external fluxes and used as constraints to solve the model by minimizing the internal fluxes. The distribution of the calculated fluxes of central metabolism were then analysed and compared with known metabolic behaviours. A striking output of the model was the energetic balance indicating that excess ATP is dissipated just before the onset of ripening, supporting the concept of the climacteric crisis (Colombi√© et al., Plant J 2014).
Importantly, both models have been cross-validated based on fluxes including the carbon-demand of the fruit. Finally, the apparent contradiction between low values of calculated fluxes compared with measured enzyme capacities suggest a complex reprogramming of the metabolic machinery during fruit development.
Bertrand Beauvoit, Sophie Colombi√©, Antoine Monier, Marie-H√©l√®ne Andrieu, Benoit Biais, Camille B√©nard, Catherine Ch√©niclet, Martine Dieuaide-Noubhani, Christine Nazaret, Jean-Pierre Mazat and Yves Gibon. 2014. Model-assisted analysis of sugar metabolism throughout tomato fruit development reveals enzyme and carrier properties in relation to vacuole expansion. Plant Cell, 26, 3222‚Äď3223.
Sophie Colombi√©, Christine Nazaret, Camille Benard, Beno√ģt Biais, Virginie Mengin, Marion Sol√©, La√ętitia Fouillen, Martine Dieuaide-Noubhani, Jean-Pierre Mazat, Bertrand Beauvoit and Yves Gibon. 2014. Modelling central metabolic fluxes by constraint-based optimization reveals metabolic reprogramming of developing Solanum lycopersicum (tomato) fruit. The Plant Journal (in press). DOI:¬†10.1111/tpj.12685.
- 10:45-11:00 : Local invited speaker (Modeling)
Speaker : Nicolas Rougier
Title : A computational mode of the somato-sensory cortex
We modeled the formation and maintenance of ordered topographic maps in the primary somatosensory cortex as well as the reorganization of representations after sensory deprivation or cortical lesion. We considered both the critical period (postnatal) where representations are shaped and the post-critical period where representations are maintained and possibly reorganized. We hypothesise that feed-forward thalamocortical connections are an adequate site of plasticity while cortico-cortical connections are believed to drive a competitive mechanism that is critical for learning. I will present main computational principles of the model as well as results and comparison with monkey data.
- 14:00-14:45 : Keynote Talk
Speaker : Gunnar Klau (CWI, Amsterdam)
Title : Finding active subnetwork modules or the Jelly Bean Mining problem
Integrative network analysis methods provide robust interpretations of high-throughput molecular profile measurements in the context of biological networks. They are used to generate novel hypotheses about the underlying cellular processes or to derive biomarkers for classification and subtyping. Of particular interest are active subnetwork modules, which are sets of genes that are responsible for the observed phenotypic changes and which are connected in a biological background network. Given appropriate scores, the problem is equivalent to finding the maximum weight connected subgraph, which is related to the prize-collecting Steiner tree problem. The talk covers different heuristic and exact approaches for solving this problem and for exploring the resulting modules by means of an application to discovering virally deregulated signaling, see http://youtu.be/LFGKek_pgjw.
- 15:05-15:20 : Local invited speaker (Data integration)
Speaker : Andreas Bikfalvi [LAMC-INSERM U1029]
Title : The use of the chick embryon model for "omics studies"
We have undertaken a major effort using the chick embryo model as a tool for cancer and angiogenesis research (Hagedorn et al PNAS 2005; Saidi Int J Cancer 2008, 2009; Cavill et al, J Prot Research, 20010; Dumartin et al., Gastroenterology, 2010; Fergelot et al, Angiogenesis 20013; Soulet et al Mol Cell Prot, 20013). More recently we have undertaken an extensive proteomic study using this model, in order to map the extracellular or membrane proteome associated with the vasculature and the stroma in an embryonic organism in-vivo. We developed a biotinylation technique in the chicken embryo combined with mass spectrometry and bioinformatic analysis. We also applied this procedure to implanted tumours growing on the chorioallantoic membrane or after induction of granulation tissue. Membrane and extracellular matrix proteins were the most abundant components identified. Relative quantitative analysis revealed differential protein expression patterns in several tissues. Through a bioinformatic approach, we determined endothelial cell protein expression signatures, which allowed us to identify several proteins not yet reported to be associated with endothelial cells or the vasculature. This is the first study reported so far that applies in-vivo biotinylation, in combination with a robust label-free quantitative proteomics approaches and bioinformatic analysis, to an embryonic organism. It also provides the first description of the vascular and matrix proteome of the embryo that may constitute the starting point for further developments. We have furthermore undertaken recently a systems biology approach using the CAM glioma model and mapped gene regulation kinetics throughout tumor development in the tumor and the stroma. We have identified major drivers of tumor development and performed a validation approach to evaluate the functional importance.
- 15:25-15:40 : Local invited speaker (Computational Biology)
Speaker :Thibault Leroy
Title : Skiing down "Mount Everest of sequencing data" : an illustration with the 1.5Gb oak genome
Throughout the northern hemisphere, oaks have important economic, societal and ecological values. Reduced costs in sequencing technologies and advances in bioinformatics over the past decade have made it possible to sequence a reference genome for pedunculate oak using a large amount of next-generation sequencing data. Data have been produced, assembled and annotated within a period of three years, reinforcing links between geneticists and computational biologists. This information will be made available to the scientific community through a genome browser. Whole genome resequencing, RNAseq and mirSeq are now being used to provide a first glimpse into gene function, regulation and evolution in a consortium where complementary approaches (population genetics, phylogeny, quantitative genetics, molecular biology, bioinformatics‚Ä¶) are considered. As an illustration, we performed a whole genome scan for genetic divergence between 4 euro! pean white oak species using more than 1.2 Tb of Pool-Seq data. This presentation will illustrate how we handle this dataset to shed lights on the molecular mechanisms underlying incipient speciation in oaks.
- 9:00-9:45 : Keynote Talk
Speaker : Peter Dominey
Title : Reservoir Computing in Higher Cognitive Functions
One of the origins of reservoir computing can be found in the primate fronto-striatal system. Local recurrent connections dominate cortico-cortical connections and thus implement the reservoir. Theprojection from cortex to the striatum is modifiable under the control of reward-related dopamine, thus forming the basis for the readout. We first specified this architecture in 1995 (J. Cog. Neuroscience) in the context of attempting to interpret prefrontal cortical neurons recorded in the behaving primate during a sensorimotor sequencing task. For the first time, a reservoir network displayed single unit activity whose distributions of different coding (sensitivity to target location, sensitivity to target rank, complex mixtures of these) was comparable to neural distributions recorded in the primate. We subsequently compared reservoir unit behavior to premotor neural behavior recorded in the primate during a task in which attention to the location of an attentional cue conditioned the proper response to an ambiguous intentional cue. Again the reservoir units displayed coding properties comparable to those in the primate cortex. More recently we have used these reservoir models to investigate the comprehension of grammatical structure in sentences. The reservoir model yields unprecedented capability to learn and generalize grammatical structures, and again, provides insight into neural coding, this time with respect to the human P600 ERP response to syntactic complexity. This research will be situated in the emerging context that suggests that cortex has fundamental reservoir properties, in particular it allows the projections of inputs into a high dimensional temporally extended space, thus providing an extreme flexibility in combinatorial coding, and the desired structure can be extracted and decoded at the readout level in the striatum. This suggests a new perspective on understanding the intermingled projections from cortex to striatum, and a new era for understanding the role of corticostriatal system in higher cognitive function, in the context of reservoir computing. Supported by FP7-ICT-612139 grant WYSIWYD.
- 10:25-10:40 : Local invited speaker (Network analysis)
Speaker : Zhanwu Dai
Title : Network analysis for studying the developmental and environmental regulation of grape quality
Grape quality is essential for the quality of resultant wine, and it is a result of complex interplay between developmental and environmental regulations. To gain a better understanding of the regulation of grape quality, metabolite, transcript, and enzyme activity profiling has been conducted over berry development and under various growth conditions. Here, we present our activity in network analysis for the rich dataset we obtained and the approaches we used to conduct our network analysis. By combining R and Cytoscape , we created correlation-based networks, and extracted interesting features of the data. Future perspectives will also be discussed.
- 10:45-11:00 : Local invited speaker (Modeling)
Speaker : Thierry Colin [IMB-INRIA]
Title : Patient specific modelling of tumor growth
The aim of this talk is to present a few results of image-based simulations of tumor growth. The idea is to provide patient-specific simulations using longitudinal sequences of images (CT-scans, MRI,...) Example concerning lung metastasis, meningiomas and liver metastasis will be given.
- 14:00-14:45 : Keynote Talk
Speaker : Hadi Quesneville
Title : Evolution et impact des √©lements transposables chez les Brassicaceae
Little is known about the evolution of the nature and impact of repeated sequences over long periods of times. The genome of several A. thaliana relatives that diverged approx. 5-40 MYA have been sequenced: Arabidopsis lyrata, Capsella rubella, Arabis alpina, Brassica rapa, Thellungiella salsuginea (formerly Thellungiella halophila), and Schrenkiella parvulum (formerly Thellungiella parvula). As well, the genomes from four other A. thaliana ecotypes have been assembled (Ler-1, Kro- 0, Bur-0, and C24). For all these genomes, we have generated a library of consensus repeat sequences that we appended to the Col-0 library ( the reference A. thalina genome) in order to compile a ‚ÄúBrassicaceae‚ÄĚ library that was used to annotate the Col-0 genome.
Our Brassicaceae TE annotations cover over 31.8 Mb of the A. thaliana genome and appear to achieve highly sensitive detection. We found that most of the copies are detected with better scores by sequences fetched from other Brassicaceae species. When measuring the coverage of TE annotations attributed to consensus sequences from each species, we found that while Col-0 contributes the largest part (36.8%), A. lyrata and B. rapa contribute as much as 17.5 and 25.8%, respectively, with modest (below 6.5%) contributions from other species and the pool of non-Col-0 A. thaliana ecotypes. The fact that most A. thaliana copies can be detected more accurately by consensus sequences built in foreign species is presumably most parsimoniously explainable by differential selective bursts among the Brassicaceae which would rule out the accelerated divergence hypothesis and indicate the long decay of these copies in A. thaliana. The present analysis therefore provides independent evidences supporting the ancient origin of the A. thaliana repeats as well as the chromosome-level distribution of old versus young copies in this species.
We found that the majority of the repeats found in the A. thaliana genome are rather ancient and likely to derive from the retention of fragments deposited by ancestral bursts that occurred early during the Brassicaceae evolution. We illustrate the way repeated sequences are composed by mutations towards genomic dark matter over time. We present results illustrating that the repeated sequences evolve in a bimodal fashion with mutation rates first dominated by the deamination of methylcytosines and then by the basal endogenous rates of transitions and transversions. Our results further suggest that the deleterious impact of repeats on gene expression as well as their regulation through small RNA-mediated pathways can last over prolonged periods. We show that a substantial pool of small RNAs corresponds to old repeats suggesting that repeat sequence divergence is accompanied by a diversifying population of small RNAs. We show that repeats have prolonged effects on the expression levels of proximal genes.
- 15:05-15:20 : Local invited speaker (Data integration)
Speaker : Rodolphe Thiébaut [ISPED/-NSERM U897]
Title : Mecanistic models for the assessment of intervention in HIV infection
I am presenting two examples of the use of simple mathematical models based on ODE for which parameters are estimated using data from HIV clinical trials evaluating either antiretroviral treatment or immune intervention (Interleukine 7 therapy).