Plant-inhabiting microorganisms interact directly with each other, forming complex microbial interaction networks. These interactions can either prevent or facilitate the establishment of new microbial species, such as a pathogen infecting the plant. Here, our aim was to identify the most likely interactions between Erysiphe alphitoides, the causal agent of oak powdery mildew, and other foliar microorganisms of pedunculate oak (Quercus robur L.). We combined metabarcoding techniques and a Bayesian method of network inference to decipher these interactions. Our results indicate that infection with E. alphitoides is accompanied by significant changes in the composition of the foliar fungal and bacterial communities. They also highlight 13 fungal operational taxonomic units (OTUs) and 13 bacterial OTUs likely to interact directly with E. alphitoides. Half of these OTUs, including the fungal endophytes Mycosphaerella punctiformis and Monochaetia kansensis, could be antagonists of E. alphitoides according to the inferred microbial network. Further studies will be required to validate these potential interactions experimentally. Overall, we showed that a combination of metabarcoding and network inference, by highlighting potential antagonists of pathogen species, could potentially improve the biological control of plant diseases.

No abstract

Amino-acid coevolution can be referred to mutational compensatory patterns preserving the function of a protein. Viral envelope glycoproteins, which mediate entry of enveloped viruses into their host cells, are shaped by coevolution signals that confer to viruses the plasticity to evade neutralizing antibodies without altering viral entry mechanisms. The functions and structures of the two envelope glycoproteins of the Hepatitis C Virus (HCV), E1 and E2, are poorly described. Especially, how these two proteins mediate the HCV fusion process between the viral and the cell membrane remains elusive. Here, as a proof of concept, we aimed to take advantage of an original coevolution method recently developed to shed light on the HCV fusion mechanism. When first applied to the well-characterized Dengue Virus (DENV) envelope glycoproteins, coevolution analysis was able to predict important structural features and rearrangements of these viral protein complexes. When applied to HCV E1E2, computational coevolution analysis predicted that E1 and E2 refold interdependently during fusion through rearrangements of the E2 Back Layer (BL). Consistently, a soluble BL-derived polypeptide inhibited HCV infection of hepatoma cell lines, primary human hepatocytes and humanized liver mice. We showed that this polypeptide specifically inhibited HCV fusogenic rearrangements, hence supporting the critical role of this domain during HCV fusion. By combining coevolution analysis and in vitro assays, we also uncovered functionally-significant coevolving signals between E1 and E2 BL/Stem regions that govern HCV fusion, demonstrating the accuracy of our coevolution predictions. Altogether, our work shed light on important structural features of the HCV fusion mechanism and contributes to advance our functional understanding of this process. This study also provides an important proof of concept that coevolution can be employed to explore viral protein mediated-processes, and can guide the development of innovative translational strategies against challenging human-tropic viruses.

To cite this version:Loic Schwaller, Stephane Robin. Exact Bayesian inference for off-line change-point detection in treestructured graphical models. Statistics and Computing, Springer Verlag (Germany), 2017, 27 (5) Abstract We consider the problem of change-point detection in multivariate time-series. The multivariate distribution of the observations is supposed to follow a graphical model, whose graph and parameters are affected by abrupt changes throughout time. We demonstrate that it is possible to perform exact Bayesian inference whenever one considers a simple class of undirected graphs called spanning trees as possible structures. We are then able to integrate on the graph and segmentation spaces at the same time by combining classical dynamic programming with algebraic results pertaining to spanning trees. In particular, we show that quantities such as posterior distributions for change-points or posterior edge probabilities over time can efficiently be obtained. We illustrate our results on both synthetic and experimental data arising from biology and neuroscience.