ABSTRACT. Exposomics researchers need to identify relevant chemicals covering the entirety of potential exposures over entire lifetimes. With over 100 million chemicals in the largest open chemical databases, coupled with broadly acknowledged knowledge gaps, researchers are faced with too much yet not enough information at the same time. Improvements in analytical technologies and computational mass spectrometry workflows coupled with the rapid growth in databases and increasing demand for high throughput “big data” services from the research community present significant challenges for both data hosts and workflow developers. This talk will showcase FAIR and Open Science developments in the Environmental Cheminformatics group, including the NORMAN Suspect List Exchange (NORMAN-SLE), MassBank, MetFrag, PubChemLite for Exposomics [1], patRoon, ShinyTPs and the Chemical Stripes. Beyond the software developments, it will showcase how these are applied in our active research projects in our data processing and integration workflows to tackle challenges in non-target exposomics studies [2,3]. The case studies will show how enhancing the FAIRness (Findability, Accessibility, Interoperability and Reusability) of open resources can mutually enhance several resources for whole community benefit. Many thanks to all group members, collaborators and colleagues who have been a part of these efforts!

References

1. Schymanski EL et al. Empowering large chemical knowledge bases for exposomics: PubChemLite meets MetFrag, Journal of Cheminformatics, 2021; 13:19, DOI: 10.1186/s13321-021-00489-0

2. Talavera Andújar B et al. (2024) Can Small Molecules Provide Clues on Disease Progression in Cerebrospinal Fluid from Mild Cognitive Impairment and Alzheimer's Disease Patients? Environ. Sci. Technol, 2024; 58(9):4181-4192. DOI: 10.1021/acs.est.3c10490

3. Talavera Andújar B et al. Exploring environmental modifiers of LRRK2-associated Parkinson’s disease penetrance: An exposomics and metagenomics pilot study on household dust. Environment International, 2024; 194:109151 DOI: 10.1016/j.envint.2024.109151

ABSTRACT. Favoured by global changes, freshwater cyanobacterial harmful blooms (HCBs) generate increasing ecological, economical and public health challenges. Microcystis, one the most pervasive genera of cyanobacteria, grows within a phycosphere, where specialized interactions with its microbiome occurs, and are suspected to influence blooms appearance and its potential toxicity. Through metagenomics, metabolomics and metabolic modelling, we characterized twelve Microcystis phycosphere cultured after isolation from a French pond. Metagenomics revealed that associated bacteria introduce new functions to the phycosphere, while functional redundancy within and across communities remains. Metabolic reaction presence in Microcystis is consistent with their genospecies, whereas community-level metabolic landscape diverges cyanobacteria’s phylogeny. On the other hand, metabolomic results lean on metabolic output led by cyanobacteria. Metabolic modelling and identification of toxic secondary metabolites biosynthetic gene cluster further highlighted differences between phycosphere metabolic capabilities and the importance of manual curation of secondary metabolism in GSMNs. These findings deepen understanding of Microcystis’ phycosphere functionning, demonstrate the relevance of multi-omics systems biology approaches, and lay the ground for further characterisation of freshwater HCB’s microbial interactions and inter-species complementarity.

ABSTRACT. Graph algorithms are essential tools for network analysis in various domains, including biology. Despite successful applications to metabolic networks, including several developments specific to these models, few implementations are openly available. Furthermore, the exchange format adopted for most genome-scale models is incompatible with the main generic graph-analysis libraries. We present Met4J, an open-source library dedicated to the structural analysis of metabolic models and their manipulation, as well as a toolbox encompassing implementations of analyses relevant to metabolism-related research. We exemplify the potential of Met4J by creating a workflow for the construction and analysis of an holobiont network. Met4J's source code, executable JAR and containers are available at https://forgemia.inra.fr/metexplore/met4j and the library artifact is accessible through the Maven central repository. High-level applications are also available on a Galaxy interface.

ABSTRACT. Genome-scale metabolic models (GEMs) are essential tools for studying metabolism, either for comparative analyses or to investigate organisms interactions. However, genome annotation, biomass formulation, and network gap-filling are key steps in constructing a relevant GEM and ensuring the biosynthesis of specialized metabolites. We present a pipeline to integrate extensive biological knowledge (genomes of closely related species, metabolic profiling studies, potential interactions with microbiota) about an eukaryotic organism in order to generate high quality GEMs. To manage genome annotation limitations, the pipeline relies on a GEM reconstruction tool that propagates annotations across closely related species through the identification of orthologous genes. It also pays particular attention to biomass formulation, using a set of metabolomic studies to create a consensus biomass composition that seeks to closely reflect biological reality, such as incorporating specialized metabolites and their precursors. The gap-filling stage of the pipeline uses a semi-automated curation process for added reactions, taking into account the presence of orthologous genes, occurrence in phylogenetically related species and potential interactions with the organism's microbiota. The final GEM applied to the brown alga Ascophyllum nodosum comprises 3,536 metabolites and 3,072 biochemical reactions, predicting the synthesis of 1,023 compounds from 38 seawater-derived metabolites. Almost all reactions (99.98%) are linked to an enzyme supported in the algal genome. This refined model provides a framework for studying host-microbiota metabolic complementarity. This pipeline offers a scalable and robust method for reconstructing high-quality GEMs in non-model eukaryotic organisms, improving metabolic network accuracy and expanding our understanding of species-specific metabolism. It also sheds lights on the various level of knowledge related to the synthesis pathways of the biomass, paving the way to future studies to be undergone.

ABSTRACT. Recent advances in k-mer indexing have facilitated the cataloging and rapid querying of planetary- scale genomic data. While these indices excel at high-throughput sequence lookups, they often lack context-rich exploration capabilities and rely on simplistic match-based queries. This gap hinders deeper investigations into variants, regulatory elements, and other features crucial for pangenomic and transcriptomic analyses. We present Vizitig, a novel system that harnesses a de Bruijn graph as the core data structure. By directly encoding overlapping k -mers from both genome and transcriptome data, Vizitig supports the processing of partially or completely unassembled sequences, making it broadly applicable from collections of genomes to eukaryotic RNA-seq. Vizitig integrates k-mer indices into a database framework, providing an intuitive, metadata-aware approach to querying. Users can select candidate regions by specific annotations (e.g., genes, motifs) or sample-specific features (e.g., abundance, presence or absence in annotated gene or a sample), retrieving relevant graph neighborhoods and associated meta-data from extensive datasets.

ABSTRACT. Over the past decade, significant efforts have been made to develop indexing solutions capable of querying sequence presence in large genomic data repositories. Recent indexing approaches have made giant steps toward the ultimate goal of indexing repositories like the SRA and ENA, leveraging k-mers for efficiency. In the case of indexing RNA samples, querying k-mer abundance is equally important than the presence itself. The current available methods for indexing abundances either fail to scale to the vast number of datasets, lose variants, or lack precision in abundance estimation. Moreover, the rapid accumulation of sequencing data presents a significant computational challenge for these structures that are mostly static. We introduce REINDEER2, a novel k-mer abundance index that addresses these limitations by providing three key properties: scalability, dynamicity, and tunable precision. REINDEER2 is highly scalable and efficient, capable of indexing thousands of RNA-seq datasets within hours while maintaining low memory usage. Unlike recent methods that sacrifice memory for completeness, REINDEER2 indexes all k-mers, ensuring nucleotide-level exploration remains possible. Additionally, it supports high-throughput queries, enabling rapid retrieval of k-mer abundance across large-scale transcriptomic datasets. One of the key advantages of REINDEER2 is its tunable abundance precision. Abundance values can be recovered with less than 1% variation from their original counts, providing reliable quantitative insights. Furthermore, REINDEER2 supports updatability: new datasets can be added efficiently without requiring a complete reindexing process through a merge operation, making it adaptable to evolving sequencing repositories. We report REINDEER2’s great efficiency at indexing collections of 1000-10,000 RNA-seq samples, and demonstrate its capacity to provide abundance estimations comparable to state-of-the-art. Availability: github.com/Yohan-HernandezCourbevoie/REINDEER2

ABSTRACT. Recent advances in high-throughput and third-generation sequencing technologies have created significant challenges in storing and managing the rapidly growing volume of read datasets. Although more than 50 specialized compression tools have been developed, employing methods such as reference-based approaches, customized generic compressors, and read reordering, many users still rely on common generic compressors (e.g., gzip, zstd, xz) for convenience, portability, and reliability, despite their low compression ratios. Here, we introduce OReO, a simple read-reordering framework that achieves high compression performance without requiring specialized software for decompression. By grouping overlapping reads together before applying generic compressors, OReO exploits inherent redundancies in sequencing data and achieves compression ratios on par with state-of-the-art tools. Moreover, because it relies only on standard decompressors, OReO avoids the need for dedicated installations and maintenance, removing a key barrier to practical adoption. We evaluated OReO on both ONT and HiFi genomic and metagenomic datasets of varying sizes and complexities. Our results demonstrate that OReO provides substantial compression gains with comparable resource usage and outperforms dedicated methods in decompression speed. We propose that future compression strategies should focus on reordering as a means to let generic compression tools fully exploit data redundancy, offering an efficient, sustainable, and user-friendly solution to the growing challenges of sequencing data storage. The OReO code is open source and available at https://github.com/girunivlille/oreo.

ABSTRACT. Madbot is a tool designed to help researchers manage and share their scientific data more easily. As research data continues to grow in volume, it becomes harder to ensure that data is accessible, reusable, and easy to understand. While other tools exist to help with parts of this process, they often lack automation, standardization, or flexibility. Madbot solves these issues by providing a simple and comprehensive solution that follows international data standards, making it easier for researchers to publish their data. It automates much of the work involved in organizing and describing data, which saves time and effort for researchers. Madbot also helps ensure that data is described correctly and consistently, following well-established standards. This makes it easier for others to find and use the data in the future. The tool connects to various global platforms like Zenodo and ENA (European Nucleotide archive), allowing researchers to submit their data directly to these repositories without hassle. Madbot’s easy-to-use interface allows users to interact with the system even if they don't have technical expertise. Behind the scenes, the tool keeps everything organized, automatically checks for mistakes, and helps researchers create accurate and high-quality metadata. Madbot’s architecture is designed to be easily extensible, enabling integration with various data storage solutions, data repositories, and metadata standards. This flexibility allows researchers to adapt the tool to their specific needs, ensuring seamless interoperability with different research infrastructures. By simplifying the process of submitting research data, Madbot encourages researchers to adopt open science principles, making their work more accessible to others. In the end, Madbot helps reduce the barriers to sharing research data and makes it easier for scientists to contribute to the global scientific community.

ABSTRACT. Reproducibility and reuse of digital bioinformatics resources are essential for the development of open and cumulative science, in line with FAIR principles. To search and reuse bioinformatics tools, scientists need to be confident enough with the reliability of their annotations. Our study focuses on the quantitative and qualitative evaluation of semantic annotations in the bio.tools registry, which serves more than 30,000 bioinformatics tool descriptions, annotated with the EDAM ontology. In this work, we propose to study how the EDAM ontology is used to categorize software based on scientific disciplines and the kind of data processing they allow. We also evaluate how qualitative the annotations are based on Shannon entropy. We emphasize that particular attention should be given to the whole set of inherited annotations from the used ontology. Our results underline the need for automatic tools to support annotation curation, reducing the annotation cost for domain experts. This study is a preliminary work aimed at designing novel annotation approaches based on the combination of knowledge graphs and large language models towards more findable and reusable bioinformatics tools.

ABSTRACT. Since 2014, a collaborative and interdisciplinary dynamic has emerged in West Africa to build lasting capacities in bioinformatics. Driven by the growing need to analyze locally produced sequencing data, this initiative has led to the development of regional infrastructures and training programs through strong partnerships between academic and research institutions, including Joseph KI-ZERBO University (UJKZ), INERA, IRD, and the LMI PathoBios. Key milestones include the establishment of bioinformatics platforms in Ouagadougou (Burkina Faso) and, more recently, in Bingerville (Côte d’Ivoire) within the WAVE-CI framework. These platforms have served as training hubs, enabling a wide range of hands-on and theoretical training—from basic GNU/Linux usage to advanced metagenomics data analysis. A major achievement of this initiative is the launch of the International Certificate in Bioinformatics and Genomics (CIBiG) in 2023–2024. This intensive program combines 154 hours of in-person courses and practical sessions with laboratory work, project-based tutoring, and personalized coaching. It covers the entire data lifecycle, from sequencing using Oxford Nanopore Technologies (ONT) to data analysis workflows including assembly, annotation, SNP detection, phylogenetics, and transcriptomic analyses. Anchored in a participatory and inclusive model, CIBiG addresses two main objectives: (1) strengthening local expertise in bioinformatics applied to agriculture and health, and (2) structuring a regional community of practice. The program is supported by committed institutional stakeholders (UJKZ, IRD, WAVE), a broad network of trainers, and a strong ambition to sustain the initiative through curriculum reforms, long-term funding strategies, and regional thematic working groups. This paper presents a ten-year retrospective on capacity-building activities, the impact of the co-constructed training programs, the pedagogical innovations used (e.g., JupyterBook, Slack, supervised internships), and the perspectives for scaling up this pioneering experience in West Africa.

ABSTRACT. Transcriptomic analysis is a powerful tool for elucidating gene expression patterns associated with specific biological conditions, offering invaluable insights into cellular responses and regulatory mechanisms. However, one major challenge in transcriptomic analysis is the need for external knowledge to interpret gene expression changes in a meaningful biological context, which can be time-consuming and prone to biases. Consequently, many gene expression signatures derived from transcriptomic data remain quite superficial and lack the depth necessary for true mechanistic understanding.

In another hand, multi-omics data allows the reconstruction of genome-scale metabolic networks which represent all biochemical reactions involved in a given organism and how these reactions interplay. Theses networks are model of phenotypic metabolism which can have many applications such as the identification of potential therapeutic targets. Nethertheless, these genome-scale metabolic model are difficult to obtain due to the tedious steps of manual curation required to obtain good quality models.

Here we introduce HUMESS (HUman Metabolis Specific Signature), a tool that seeks to bridge the gap between both approaches. Using a snakemake implementation, HUMESS integrates quantitative transcriptomic data with metablic network modeling by (i) identifying significantly expressed genes from quantitative 3'seq-RNA Profiling (3'SRP) sequencing data, and (ii) uses a modified version of CarveMe - top-down approach for metabolic model reconstruction from a universal human metabolic model - for building a metabolic model specific to the gene differentially expressed. The metabolic model is then (iii) extensively analysed for identifying reactions essential to sustain the human metabolic phenotype.

In order to facilitate the analysis of the results, an online Rshiny interface has been developed allowing an in depth exploration of the results. The demo will show how this interface has successfully been used to analyze various stages of human embryonic stem cell development as described in the HUMESS preprint paper

ABSTRACT. With the increasing accessibility of genome sequencing, a crucial step now involves comparing the sequenced genomes of the same species to identify structural variations and syntenic regions. The SyRI(1) tool currently enables such analyses, while plotsr(2) provides a graphical representation of the results. However, this representation remains static and could be enhanced for a more interactive exploration of the data. We introduce SynFlow, a web application designed to provide a dynamic and interactive visualization of SyRI analysis results. Featuring an intuitive and responsive interface, SynFlow allows users to explore genome alignments and structural variations. The application includes interactive features such as zooming, panning, and filtering of bands based on type and length, thereby facilitating the analysis and interpretation of genomic data. SynFlow also enhances accessibility and usability by integrating real-time interactions and supports the simultaneous visualization of up to ten genomes, making it a powerful tool for researchers in comparative genomics. Code is available at: https://github.com/SouthGreenPlatform/SynFlow

ABSTRACT. The use of reference genomes introduces a bias to all genomic studies that rely on them, since a single individual from a population is not representative of the full genetic diversity. Pangenomes, accessible thanks to lower sequencing costs, bring together several complete genomes in a single data structure. A compact way to represent this complex data is the pangenome graph, which groups similar or divergent regions of the graph into nodes that may or may not be traversed by their individuals genomes. Many tools output a graph in GFA format, but there is a lack of tools to manipulate them. The current tools that manipulate graphs are VG or ODGI, but they require specific formats, can be time consuming, and do not allow as many different manipulations as Gratools in a single tool.

Gratools is an efficient tool for manipulating pangenome graphs in GFA format. It has several features that I'll demonstrate in a demo for graph description, extraction and analysis. Gratools begins with a one-time indexing step to extract and store essential information in BAM and BED files, it only need to be performed once per graph and significantly accelerates downstream analyses performed by GraTools.

The main steps of the demo will be to list the samples and chromosomes present in the GFA file and their sizes using the list_sample and list_chr commands. Second, a general assessment of the depth of the nodes with the depth_nodes command to get a first idea of the distribution of the nodes in this file and nodes can be filtered by size.

Next, the core_dispensable_ratio command is used to analyze the core genome and the dispensable genome according to user-defined limits. The get_segments_by_depth command is used to list nodes according to the number of individuals in which they are present (for example, you can request the list of nodes that make up the core genome). Finally, we'll show you how to extract a subgraph using the extract_sub_graph command.

ABSTRACT. Abstract Transposable elements (TEs) play a major role in the structure and evolution of eukaryote genomes. Thanks to their ability to move around and to replicate within genomes, they are probably the most important contributors to genome plasticity. Their detection and annotation are now considered mandatory to any genome sequencing project. The REPET package [1, 2] integrates bioinformatics pipelines dedicated to detect, annotate and analyze TEs in genomics sequences. It includes two main pipelines : (i) TEdenovo, that search for interspersed repeats, build consensus sequences and classify them according to TE features [3] and (ii) TEannot, which mines a genome with a library of TE sequences, for instance the one produced by the TEdenovo pipeline, to provide TE annotations. With the latest version, REPET has evolved beyond a simple downloadable archive that required manual dependency management. It is now a Snakemake pipeline, offering a clearer workflow, easier installation, and better compatibility with up-to-date tools. These improvements significantly reduce installation and configuration challenges, making REPET more accessible and efficient. In this demo, we will present the REPET strategy for identifying TEs, highlight the key improvements and new features introduced in the latest version, and demonstrate REPET on a small dataset. By the end of the session, attendees will have acquired the basic knowledge needed to detect and annotate TEs in genomes.

References 1. Flutre T, Duprat E, Feuillet C, Quesneville H (2011) Considering Transposable Element Diversification in De Novo Annotation Approaches. PLoS ONE 6(1): e16526. https://doi.org/10.1371/journal.pone.0016526 2. Quesneville H, Bergman CM, Andrieu O, Autard D, Nouaud D, Ashburner M, et al. (2005) Combined Evidence Annotation of Transposable Elements in Genome Sequences. PLoS Comput Biol 1(2): e22. https://doi.org/10.1371/journal.pcbi.0010022 3. Hoede C, Arnoux S, Moisset M, Chaumier T, Inizan O, Jamilloux V, et al. (2014) PASTEC: An Automatic Transposable Element Classification Tool. PLoS ONE 9(5): e91929. https://doi.org/10.1371/journal.pone.0091929