Cette thèse porte sur la mise en place et le développement d’une approche expérimentale pour l’étude de la dynamique spatio-temporelle de réseaux de réactions à base d’ADN. Nos résultats démontrent la capacité des réseaux d’ADN à se spatialiser sous la forme d’ondes progressives. Nous avons également pu obtenir des motifs stationnaires à base d’ADN et d’assemblages de billes. Ce travail contribue donc à la conception de motifs spatio-temporels de réactions chimiques et de matériaux par le biais de réseaux réactionnels biochimiques programmables. Nous apportons également de nouvelles données sur l’émergence d’ordre spatio-temporel à partir de processus de réaction-diffusion. De ce fait, cette étude contribue à une meilleure compréhension des principes fondamentaux qui régissent l’apparition d’une auto-organisation moléculaire dans un système chimique hors-équilibre. De plus, la combinaison de réseaux synthétiques d’ADN, du contrôle du coefficient de diffusion de plusieurs espèces d’ADN et de la micro-fluidique peut donner lieu à des motifs spatiaux stables, comme par exemple, les fameuses structures de Turing, ce qui tend à confirmer le rôle de celles-ci dans la morphogénèse. / This PhD work is devoted to developing an experimental framework to investigate chemical spatiotemporal organization through mechanisms that could be at play during pattern formation in development. We introduce new tools to increase the versatility of DNA-based networks as pattern-forming systems. The emergence of organization in living systems is a longstanding fundamental question in biology. The two most influential ideas in developmental biology used to explain chemical pattern formation are Wolpert's positional information and Turing's reaction-diffusion self-organization. In the case of positional information, the pattern emerges from a pre-existing morphogen gradient across space that provides positional values as in a coordinate system. Whereas, the Turing mechanism relies on self-organization by driving a system of an initially homogeneous distribution of chemicals into an inhomogeneous pattern of concentration by a process that involves solely reaction and diffusion. Although numerical simulations and mathematical analysis corroborate the incredible potential of reaction-diffusion mechanisms to generate patterns, their experimental implementation is not trivial. And despite of the exceptional achievements in pattern formation with Belousov–Zhabotinsky systems, these are difficult to engineer, thus limiting their experimental implementation to few available mechanisms. In order to engineer reaction-diffusion systems that display spatiotemporal dynamics the following three key elements must be controlled: (i) the topology of the network (how reactions are linked to each other, i.e. in a positive or negative feedback manner), (ii) the reaction rates and (iii) the diffusion coefficients. Recently, using nucleic acids as a substrate to make programmable dynamic chemical systems together with the lessons from synthetic biology and DNA nanotechnology has appeared as an attractive approach due to the simplicity to control reaction rates and network topology by the sequence. Our experimental framework is based on the PEN-DNA toolbox, which involves DNA hybridization and enzymatic reactions that can be maintained out of equilibrium in a closed system for long periods of time. The programmability and biocompatibility of the PEN-DNA toolbox open new perspectives for the engineering of the reaction-diffusion chemical synthesis, in particular in two directions. Firstly, to study biologically-inspired pattern-forming mechanisms in simplified, yet relevant, experimental conditions. Secondly to build new materials that would self-build by a process inspired from embryo morphogenesis. We worked towards the goal of meeting the two requirements of Turing patterning, transferring chemical spatiotemporal behavior into material patterns, and imposing boundary conditions to spatiotemporal patterns. Therefore, the structure of this document is divided into four specific objectives resulting in four chapters. In chapter 1 we worked on testing a DNA-based reaction network with an inhibitor-activator topology. In chapter 2 we focused on developing a strategy to tune the diffusion coefficient of activator DNA strands. In chapter 3 we explored how chemical patterns determine the shape of a material. Finally, in chapter 4 we addressed the issue of controlling the geometry over a DNA-based reaction-diffusion system. Overall, we have expanded the number of available tools to study chemical and material pattern formation and advance towards Turing patterns with DNA.

This paper investigates the use of deep Convolutional Neural Networks for modeling the intronic regulation of Alternative Splicing on the basis of DNA sequence. By training the CNN on massively parallel synthetic DNA libraries of Alternative 5'-splicing and Alternatively Skipped exon events, the model is capable of predicting the relative abundance of alternatively spliced mRNA isoforms on held-out library data to a very high accuracy (R2 = 0.77 for Alt. 5'-splicing). Furthermore, the CNN is shown to generalize alternative splicing across cell lines efficiently. The Convolutional Neural Net is tested against a Logistic regression model and the results show that while prediction accuracy on the synthetic library is notably higher compared to the LR model, the CNN is worse at generalizing to new intronic contexts. Tests on non-synthetic human SNP genes suggest the CNN is dependent on the relative position of the intronic region it was trained for, a problem which is alleviated with LR. The increased library prediction accuracy of the CNN compared to Logistic regression is concluded to come from the non-linearity introduced by the deep layer architecture. It adds the capacity to model complex regulatory interactions and combinatorial RBP effects which studies have shown largely affect alternative splicing. However, the architecture makes interpreting the CNN hard, as the regulatory interactions are encoded deep within the layers. Nevertheless, high-performance modeling of alternative splicing using CNNs may still prove useful in numerous Synthetic biology applications, for example to model differentially spliced genes as is done in this paper. / Den här uppsatsen undersöker hur djupa neurala nätverk baserade på faltning ("Convolutions") kan användas för att modellera den introniska regleringen av Alternativ Splicing med endast DNA-sekvensen som indata. Nätverket tränas på ett massivt parallelt bibliotek av syntetiskt DNA innehållandes Alternativa Splicing-event där delar av de introniska regionerna har randomiserats. Uppsatsen visar att nätverksarkitekturen kan förutspå den relativa mängden alternativt splicat RNA till en mycket hög noggrannhet inom det syntetiska biblioteket. Modellen generaliserar även alternativ splicing mellan mänskliga celltyper väl. Hursomhelst, tester på icke-syntetiska mänskliga gener med SNP-mutationer visar att nätverkets prestanda försämras när den introniska region som används som indata flyttas i jämförelse till den relativa position som modellen tränats på. Uppsatsen jämför modellen med Logistic regression och drar slutsatsen att nätverkets förbättrade prestanda grundar sig i dess förmåga att modellera icke-linjära beroenden i datan. Detta medför dock svårigheter i att tolka vad modellen faktiskt lärt sig, eftersom interaktionen mellan reglerande element är inbäddat i nätverkslagren. Trots det kan högpresterande modellering av alternativ splicing med hjälp av neurala nät vara användbart, exempelvis inom Syntetisk biologi där modellen kan användas för att kontrollera regleringen av splicing när man konstruerar syntetiska gener.