MeSU Users Event 2024

MeSU platform is currently switching from its old supercomputer to its new one. At the same time, the platform is welcoming more and more users from different scientific domains. In this keynote, we will report on the current state and usage of the platform, and describe ongoing evolution projects.

The planetary climate conditions in the solar system highlight certain common features, modulated by the specific characteristics of each planet. These different conditions provide full-scale “laboratories” for studying phenomena in their entirety. There are three main study tools: in situ measurements, laboratory measurements and numerical modeling. At IPSL, complex climate models are developed. Derived from terrestrial models, these models are applied to terrestrial planets such as Mars and Venus, to gas giants and now to exoplanets. I'll be presenting the uses of these tools with MeSU.

The large-scale structures of the Universe provide essential clues to its composition and evolution. By studying the evolution of these structures, we gain insights into various properties of the Universe, including the amount of dark matter, dark energy and baryonic matter. One probe of these structures is the thermal Sunyaev-Zel’dovich (tSZ), which depends on the Universe's composition. In this talk, I will present the simulations I have developed to investigate how dark energy influences the tSZ effect.

Hybrid electrolytes, such as binary mixtures of organic solvent and water, are very promising as they offer a unique opportunity to harness water as the oxygen or proton source in electrosynthesis. In this talk, combining molecular dynamics with experiments, I will show that modulating water-organic solvent interactions allows to control the size and composition of aqueous domains forming in these hybrid electrolytes. For reactions proceeding first via the activation of water, such as hydrogen evolution reaction, the fine tuning of the aqueous domains impacts the kinetics of the reaction. Instead, for organic substrates reacting prior to water, aqueous domains have no impact on the reaction kinetics, while selectivity may be affected. Our results open new perspectives for using hybrid electrolytes to optimize electrosythetic reactions.

In the recent theoretical work [arXiv:2409.11773], I study the mechanical properties of polymerized membranes, specifically the renormalization of their elastic properties. My main goal is to investigate the anomalous elasticity induced by small thermal fluctuations in a general membrane model, with direct applications to the elastic properties of graphene sheets, cell cytoskeletons, or even sheets of paper. To this end, I employ a modern version of statistical mechanics known as statistical field theory, which provides powerful tools, primarily perturbative theory and Feynman diagrams. In this framework, any physical observable can be evaluated with arbitrary precision by computing increasingly difficult integrals; the more integrals included, the more accurate the result. In this work, I push for high precision in the computation of anomalous elasticity and have computed 57 million integrals, analytically. This requires the use of a supercomputer to manipulate very large polynomials and run several dedicated software programs, originally designed for particle physics.

With intensive Monte Carlo simulations and finite size scaling we undertake a percolation analysis of Wegner's three-dimensional lattice gauge model. We confirm that the loops threading excited plaquettes percolate at the thermal critical point and we show that their critical exponents coincide with the ones of the loop representation of the dual 3D Ising model. We then construct Fortuin-Kasteleyn (FK) clusters using a random-cluster representation and find that they also percolate at and moreover give access to all thermal critical exponents. The Binder cumulants of the percolation order parameter of both loops and FK clusters demonstrate a pseudo first order transition. This study sheds light on the critical properties of pure lattice gauge theories, and consequently, quantum error correction.

The human microbiome corresponds to the whole community of microorganisms that live inside various sites of the human body. The gut microbiome, especially, is known to be associated with various health outcomes, notably the inflammatory bowel disease or IBD. Our team uses shotgun metagenomics sequencing to characterize the full DNA content of the microbiome in an unbiased way. Because each of the numerous samples can yield tens of millions of sequences, we use bioinformatics approaches to study their content and provide insights about the human gut health. More precisely, in the suivitheque and suivitheque oral projects, I profiled the bacterial species composition as well as individual bacterial strains in microbiomes.

The inability to produce functional antibodies is responsible for increased susceptibility to infections and non-infectious complications in subjects with Primary Antibody Deficiencies (PAD). While the biomarkers of this defect have not been well elucidated, we propose the virome as a potential component of the microbiome that can segregate between common PAD clinical phenotypes. After filtering and concentrating for virus-like particles from stool samples of 31 healthy adults and 48 subjects with PAD, we performed viral sequencing after which resulting sequencing data was analyzed on the MeSU platform. This involved using in-house containerized Nextflow-based pipelines as well as existing viromics workflows for sequence genome assembly, viral contig identification, and bacterial host prediction. We observed a significant increase in the abun- dance of viral families such as Lipothrixviridae and Straboviridae in PAD patients as compared to healthy donors. Furthermore, random forest classification showed that virome data could discriminate between Control and PAD patients with a precision of upto 92%. These virome alterations that are specific to PAD groups as well as virome-based machine learning classification presents the viral community as a potential biomarker to differentiate between PAD phenotypes.

SUMMIT is a service unit of Sorbonne University. It provides researchers and companies with specialized engineering profiles and takes charge of development activities. In addition, SUMMIT offers privileged access to Sorbonne Université's research and development (R&D) ecosystem. SUMMIT is organized into three technical departments: Engineering and Systems, Mathematics and Applications, and Observatory and Society.

LabRADoR, Lab de Ressources et d'Accompagnement aux Données de la Recherche, is a pilot project designed to answer your questions about research data, whether you need to open or protect it. Supported by the SACADO service unit and the Sorbonne University Library (BSU), this portal aims to address your questions in all their complexity and various dimensions (technical, legal, documentary), drawing on the skills of several university structures: the Research and Innovation Department, the Research and Development Department, the IT Services Department, the Legal and Institutional Affairs Department and the Data Protection Officer.

MeSU platform has changed a lot since its inception and has been growing steadily year after year. In this keynote, we will review a couple of past milestones, report on the current state and usage of the platform, and describe ongoing evolution projects.

Plastic debris is known to accumulate at specific regions at sea such as the Great Pacific Garbage Patch. However, this is not the case in many basins worldwide, including the Mediterranean Sea, heavily polluted. In this project, we decided to identify the locations where several plastic debris passes, rather than look for locations where it accumulates. To this aim, we needed to simulate the fate of ~150M virtual plastic particles, in order to analyse their trajectories. To achieve such a great computational challenge, we used TrackMPD model (Matlab language) and exploited the beta calculator of Mesu. The calculation was split over 96 cores of 4 different nodes. PBS scripts were used to run consecutively the different phases of plastic lifetime: their release at sea, their trajectory calculation (transport by current, beaching, etc.), the identification of the regions where most of debris passes, and other statistics. The prompt assistance of the mesu staff all along the different stages of the analyses was crucial for the success of the work. The results were published in two publications, one on Nature Communications, and the other in Environmental Science&Technology.

In marine sciences, environmental variables such as hydrological and physical variables are essential to better understand species distribution. These data can be cumbersome and time-consuming to process. Here, we studied variations in plankton distribution between night and day to better understand ecosystem functioning. To make the link with environmental conditions, we needed to separate day and night data based on 64 years of hourly environmental variables. The MeSU platform offered a great support to learn how to use the supercomputer and optimise processing time.

LAMS research works focuses on cultural heritage objects and techniques from a physico-chemical point of view. Among these, dyeing materials are important due to their wide range of uses and their presence all over the world and in different time periods. In particular, we have worked on a famous embroidery "la tapisserie de Bayeux", which has undergone photochemical degradation leading to loss of information or colour fading. To study these systems computational chemistry can be of interest to understand the stability or reproduce the colour of these dyed objects.
The first part of this talk will present different methods of computational chemistry used on MeSU cluster and discuss the first results obtained.
The second part of this talk will focus on one of the most renowned dye plant in Europe, and particularly in South of France, Madder. Madder is the origin of the famous “Adrianople red” (or red Turkish) giving a bright red on cotton, in a long and tedious recipe. Its dyes are extracted from its roots, with Alizarin and Purpurin being the most important. Prior to the advent of computational chemistry, even though researchers worked diligently, deciphering the experimental spectroscopic properties of Madder's colorants posed a persistent challenge. This was primarily due to the extraction difficulty, impurities, the instability of some compounds, and their high price. Now, we can overcome this challenge with a new approach. Thanks to the application of computational spectroscopy techniques, we can reproduce, understand and predict the spectra of compounds that were previously unknown or inaccessible experimentally.

As scientific datasets become more intricate and larger in size, a need for advanced data analysis algorithms arises, to enable efficient visualization and exploration. Topological Data Analysis (TDA) is a set of techniques for extracting, measuring and comparing structural features and patterns of datasets using high-level topological abstractions such as critical points, Morse-Smale complexes or persistence diagrams. The Topology Tool Kit (TTK) is a C++ open-source project that provides an efficient and robust implementation of key algorithms for TDA. However, as datasets grow larger, the computation of topological abstractions may become unreasonably costly in time and may even exceed the capacity of a single commodity computer’s main memory. These challenges can be addressed by using distributed-memory systems and leveraging their significantly larger capacities. This was accomplished by porting TTK to a distributed computation model, using MPI (Message Passing Interface). In our presentation, we will show how we have used MeSU-beta, specifically through benchmarks of several algorithms of our distributed-memory approaches. Furthermore, we will present the computation of a real-life use case of an analysis pipeline combining several algorithms on the largest dataset publicly available that we have found (120 billion vertices) on 64 nodes of 24 cores each (for a total of 1,536 cores).

The non-covalent interaction (NCI) index is a formalism that allows to identify and classify regions of non-covalent interactions in molecules. The implementation of this method in NCIPLOT permits to perform calculations for both small and large molecules, opening the possibility to use it in the study of systems of biological interest. However, as it is, NCIPLOT fails to become a routinely-used tool in the sectors of the scientific community that are interested in such an analysis, primarily originating from the field of biology, because it requires a user that is familiar with command-line- based computational tools. Here, we present NCIWEB (https://nciweb.dsi.upmc.fr), a web server that permits to launch NCIPLOT calculations in MeSU directly from the browser, simply by providing the initial structure of the molecule(s) or its ID number in the PDB database; and interactively choosing certain options for the type of analysis to be performed. Once the user has uploaded all the information needed for the calculation, the data is transferred to MeSU, where an NCIPLOT job is launched. This process is automatized, and the user is informed of its progress via email, being notified when the job starts running, and when results are available. The NCIWEB server was presented in a scientific article in August 2023 [J. Chem. Inf. Model. 2023, 63, 15, 4483– 4489], and has received a warm reception within the scientific community, which has demonstrated a keen interest in this web-based tool.

My intervention, during which I will represent the team of the ANR Project « PhiLat – Lexique philosophique de la latinité », aims at giving our usage feedback on the services that SACADO provides us in the framework of the MeSU platform, viz., data storage, hosting, and exploitation of a virtual machine. In this regard, in the first place, I will briefly introduce our project and describe the service of the MeSU platform from which we benefit, focusing on their technical features (size of storable data, operating systems available, RAM provided, etc.) and on the warranties offered by SACADO (in terms of service continuity, problem-solving time and strategies, data security and integrity, etc.). I will then give a report on our experience with these services and with the management of blackouts by the staff of SACADO. Building on this experience, as well as on the comparison of the offer by SACADO with others present on the market, I will finally give our general assessment of the aforesaid services and our degree of satisfaction with them.

Research funders now require research data to be opened up, where possible, at the end of projects. This means anticipating the data management practices that will be put in place. Whether it is securing data to keep it closed because of planned patents, or opening it because it is mandatory for geographic data, planning ahead is key. The Research Data and Digital Humanities team can provide advice and support to help you plan your project's data management.