AbstractThe prediction of a protein structure is important for studying its function at the molecular level. However, prediction of the structural fold from the protein’s amino acid sequence is one of the most challenging problems in computational biology. Protein structure prediction based on comparative modelling is the most efficient way to produce structural models when it can be performed. A strategy called Nano-ORION use the concept of Protein Units (PU), small and compact structure… Show more AbstractThe prediction of a protein structure is important for studying its function at the molecular level. However, prediction of the structural fold from the protein’s amino acid sequence is one of the most challenging problems in computational biology. Protein structure prediction based on comparative modelling is the most efficient way to produce structural models when it can be performed. A strategy called Nano-ORION use the concept of Protein Units (PU), small and compact structure fragments, as structural templates. This approach showed promising results during the last CASP13 protein structure prediction competition, but lacked specificity. In this work, we introduce a novel approach using deep learning called Deep-Nano-ORION. This new strategy aims at developing a deep neural network to decipher sequence-structure relationships between PUs sharing high structural similarity. The model is then able to predict whether or not a given PU is a suitable local template. Using this model, Deep-Nano-ORION re-ranked the Nano-ORION alignments, increasing its specificity. Show less

Abstract:Untargeted metabolomics generates a large amount of data. Software packages already exist for LC-MS and GC-MS data analysis and are available inside automated data analysis workflows, hosted in platforms such as Workflow4Metabolomics (W4M). The R/Bioconductor open-source package XCMS was specifically developed for the preprocessing steps, consisting among others in peak picking, retention time correction and grouping. The selection of parameters is crucial in order to yield the… Show more Abstract:Untargeted metabolomics generates a large amount of data. Software packages already exist for LC-MS and GC-MS data analysis and are available inside automated data analysis workflows, hosted in platforms such as Workflow4Metabolomics (W4M). The R/Bioconductor open-source package XCMS was specifically developed for the preprocessing steps, consisting among others in peak picking, retention time correction and grouping. The selection of parameters is crucial in order to yield the best results. The preprocessing analysis depends on the analytical setup of a study (LC-MS instruments, methods, etc.). Which is why the XCMS parameters must be optimized to fit the context of the processed study. IPO (Isotopologue Parameter Optimization) is an R/Bioconductor package for the automated XCMS parameters optimization. This report presents how the W4M prototyped version of IPO was upgraded and integrated into PhenoMeNal, the European standardized e-infrastructure for large-scale computing of medical metabolomics, and run on several metabolomics studies from the Metabolights database. Preliminary results from the large-scale processing of all public Metabolights studies show indeed differences between optimized parameters and default values suggested by XCMS. Show less

Abstract:Gas chromatography coupled with mass spectrometry, or GC-MS, is a commontechnique for metabolites detection and identification. They are small moleculesstudied in metabolomics, a new and rapidly developing study field. A specializeddatabase is necessary in order to annotate the mass spectra to identify metabolitesof interest. This report describes how GC-MS data are generated and details theprograms and methods used to develop an annotation tool conceived to… Show more Abstract:Gas chromatography coupled with mass spectrometry, or GC-MS, is a commontechnique for metabolites detection and identification. They are small moleculesstudied in metabolomics, a new and rapidly developing study field. A specializeddatabase is necessary in order to annotate the mass spectra to identify metabolitesof interest. This report describes how GC-MS data are generated and details theprograms and methods used to develop an annotation tool conceived to comparemass spectra with the Golm database, which is based in Germany.Sujet du stage :Développement d'un outil en Perl Orienté Objet (POO) intégré au système de gestion de workflows bioinformatiques Galaxy pour l’analyse de données issues de Spectrométrie de Masse couplée à de la Chromatographie en phase Gazeuse (GC-MS) au sein de la Plateforme d’Exploration du Métabolisme de l’INRA de Clermont Ferrand-Theix.Description:Développement d’un outil d’annotation de données GC-MS exploitant la base de données référence Golm (Golm Metabolome Database). Cet outil sera intégré à la plateforme en ligne collaborative et gratuite Galaxy, dans le cadre du projet français Workflow4Metabolomics qui vise à développer une chaîne complète de traitement de données métabolomiques. Show less