3.1.1 Context

The scientific discovery process is very often based on cycles of measurement, classification, and generalisation. It is easy to argue that this is especially true in the biological sciences. The proteins that exist today represent the molecular product of some three billion years of evolution. Therefore, comparing protein sequences and structures is important for understanding their functional and evolutionary relationships 59, 79. There is now overwhelming evidence that all living organisms and many biological processes share a common ancestry in the tree of life. Historically, much of bioinformatics research has focused on developing mathematical and statistical algorithms to process, analyse, annotate, and compare protein and DNA sequences because such sequences represent the primary form of information in biological systems. However, there is growing evidence that structure-based methods can help to predict networks of protein-protein interactions (PPIs) with greater accuracy than those which do not use structural evidence 62, 85. Therefore, developing techniques which can mine knowledge of protein structures and their interactions is an important way to enhance our knowledge of biology 49.

3.1.2 Formalising and Exploiting Domain Knowledge

Concerning protein structure classification, we aim to explore novel classification paradigms to circumvent the problems encountered with existing hierarchical classifications of protein folds and domains 44, 45, 80. In particular it will be interesting to set up fuzzy clustering methods taking advantage of our previous work on gene functional classification 54, but instead using Kpax domain-domain similarity matrices. A non-trivial issue with fuzzy clustering is how to handle similarity rather than mathematical distance matrices, and how to find the optimal number of clusters, especially when using a non-Euclidean similarity measure. We will adapt the algorithms and the calculation of quality indices to the Kpax similarity measure. More fundamentally, it will be necessary to integrate this classification step in the more general process leading from data to knowledge called Knowledge Discovery in Databases (KDD) 57.

For example, domain family classification 48, 65 is relevant for studying domain-domain interactions (DDI). Our previous work on Knowledge-Based Docking (KBDOCK, 4, 6) will be updated and extended using newly published DDIs. Methods for inferring new DDIs from existing protein-protein interactions (PPIs) will be developed. Efforts should be made for validating such inferred DDIs so that they can be used to enrich DDI classification and predict new PPIs.

In parallel, we also intend to design algorithms for leveraging information embedded in biological knowledge graphs (also known as complex networks). Knowledge graphs mostly represent PPIs, integrated with various properties attached to proteins, such as pathways, drug binding or relation with diseases 69. Setting up similarity measures for proteins in a knowledge graph is a difficult challenge. Our objective is to extract useful knowledge from such graphs in order to better understand and highlight the role of multi-component assemblies in various types of cell or organisms. Ultimately, knowledge graphs can be used to model and simulate the functioning of such molecular machinery in the context of the living cell, under physiological or pathological conditions.

3.1.3 Function Annotation in Large Protein Graphs

Knowledge of the functional properties of proteins can shed considerable light on how they might interact. However, huge numbers of protein sequences in public databases such as UniProt/TrEMBL lack any functional annotation, and the functional annotation of such sequences is a highly challenging problem. We are developing graph-based and machine learning techniques to annotate automatically the available unannotated sequences with functional properties such as Enzyme Commission (EC) numbers and Gene Ontology (GO) terms (note that these terms are organised hierarchically allowing generalization/specialization reasoning). The idea is to transfer annotations from expert-reviewed sequences present in the UniProt/SwissProt database (about 560 thousands entries) to unreviewed sequences present in the UniProt/TrEMBL database (about 80% of 180 millions entries). For this, we have to learn from the UniProt/SwissProt database how to compute the similarity of proteins sharing identical or similar functional annotations. Various similarity measures can be tested using cross-validation approches in the UniProt/SwissProt database. For instance, we can use primary sequence or domain signature similarities. More complex similarities can be computed with graph-embedding techniques.

This work constituted Bishnu Sarker's PhD thesis (defended in April 2021) 37 with a major contribution called GrAPFI (Graph-based Automatic Protein Function Inference) and is now continued in Md Kamrul Islam's thesis (ongoing since 2019).

3.2.1 Context

At the molecular level, each biomolecular interaction is embodied by a physical 3D interface. Therefore, if the 3D structures of a pair of interacting protein/NA molecules are known, it should in principle be possible for a docking algorithm to use this knowledge to predict the structure of the complex. However, modeling protein and even more RNA flexibility accurately during docking is very computationally expensive. This is due to the very large number of internal degrees of freedom in each molecule, associated with twisting motions around covalent bonds. Therefore, it is highly impractical to use detailed force-field or geometric representations in a brute-force docking search. Instead, most docking algorithms use fast heuristic methods to perform an initial rigid-body search in order to locate a relatively small number of candidate binding orientations, and these are then refined using a more expensive interaction potential or force-field model, which might also include flexible refinement using molecular dynamics (MD), for example.

3.2.2 Polar Fourier Docking Correlations

In our Hex protein docking program 71, the shape of a protein molecule is represented using polar Fourier series expansions of the form

where $\sigma \left(\underline{x}\right)$ is a 3D shape-density function, $a_{nlm}$ are the expansion coefficients, $R_{nl}\left(r\right)$ are orthonormal Gauss-Laguerre polynomials and $y_{lm}(\theta ,\varphi )$ are the real spherical harmonics. The electrostatic potential, $\varphi \left(\underline{x}\right)$, and charge density, $\rho \left(\underline{x}\right)$, of a protein may be represented using similar expansions. Such representations allow the in vacuo electrostatic interaction energy between two proteins, A and B, to be calculated as 61

This equation can be demonstrated using the notion of overlap between 3D scalar quantities to give a physics-based scoring function. If the aim is to find the configuration that gives the most favourable interaction energy, then it is necessary to perform a six-dimensional search in the space of available rotational and translational degrees of freedom. By re-writing the polar Fourier expansions using complex spherical harmonics, we showed previously that FFT techniques may be used to accelerate the search in up to five of the six degrees of freedom 72. Furthermore, we also showed that such calculations may be accelerated dramatically on modern graphics processor units 8, 12. Consequently, we are continuing to explore new ways to exploit the polar Fourier approach.

3.2.3 Assembling Symmetrical Protein Complexes

Although protein-protein docking algorithms are improving 63, 73, it still remains challenging to produce a high resolution 3D model of a protein complex using ab initio techniques. This is mainly due to the problem of structural flexibility described above. However, with the aid of even just one simple constraint on the docking search space, the quality of docking predictions can improve considerably 12, 72. In particular, many protein complexes involve symmetric arrangements of one or more sub-units, and the presence of symmetry may be exploited to reduce the search space considerably 47, 70, 78. For example, using our operator notation (in which $\widehat{R}$ and $\widehat{T}$ represent 3D rotation and translation operators, respectively), we have developed an algorithm which can generate and score candidate docking orientations for monomers that assemble into cyclic ($C_n$) multimers using 3D integrals of the form

where the identical monomers A and B are initially placed at the origin, and ${\omega }_n=2\pi /n$ is the rotation about the principal $n$-fold symmetry axis. This example shows that complexes with cyclic symmetry have just 4 rigid body degrees of freedom (DOFs), compared to $6(n-1)$ DOFs for non-symmetrical $n$-mers. We have generalised these ideas in order to model protein complexes that crystallise into any of the naturally occurring point group symmetries ($C_n$, $D_n$, $T$, $O$, $I$). This approach was published in 2016 11, and was subsequently applied to several symmetrical complexes from the “CAPRI” blind docking experiment 56. Although we currently use shape-based FFT correlations, the symmetry operator technique may equally be used to build and refine candidate solutions using a more accurate coarse-grained (CG) force-field scoring function.

3.2.4 Coarse-Grained Models

Many approaches have been proposed in the literature to take into account protein (and more recently RNA/DNA) flexibility during docking. The most thorough methods rely on expensive atomistic simulations using MD. However, much of a MD trajectory is unlikely to be relevant to a docking encounter unless it is constrained to explore a putative protein-protein/NA interface. Consequently, MD is normally only used to refine a small number of candidate rigid body docking poses. A much faster, but more approximate method is to use "coarse-grained" (CG) normal mode analysis (NMA) techniques to reduce the number of flexible degrees of freedom to just one or a handful of the most significant vibrational modes 55, 64, 66, 68. In our experience, docking ensembles of NMA conformations do not give much improvement over basic FFT-based soft docking 83, and it is very computationally expensive to use side-chain repacking to refine candidate soft docking poses 5.

In the last few years, CG force-field models have become increasingly popular in the MD community because they allow very large biomolecular systems to be simulated using conventional MD programs 46. Typically, a CG force-field representation replaces the atoms in each amino acid with from 2 to 4 “pseudo-atoms”, and it assigns each pseudo-atom a small number of parameters to represent its chemo-physical properties. By directly attacking the quadratic nature of pair-wise energy functions, coarse-graining can speed up MD simulations by up to three orders of magnitude. Nonetheless, such CG models can still produce useful models of very large multi-component assemblies 77. Furthermore, this kind of CG model effectively integrates out many of the internal DOFs to leave a smoother but still physically realistic energy surface 60. We are currently developing a CG scoring function for RNA-protein docking by fragments assembly. This work is part of the PhD project of Anna Kravchenko.

3.2.5 Assembling Multi-Component Complexes and Integrative Structure Modelling

We also want to develop related approaches for integrative structure modeling using cryo-electron microscopy (cryo-EM). Thanks to recent developments in cryo-EM instruments and technologies, it is now feasible to capture low resolution images of very large macromolecular machines. However, while such developments offer the intriguing prospect of being able to trap biological systems in unprecedented levels of detail, there will also come with an increasing need to analyse, annotate, and interpret the enormous volumes of data that will soon flow from the latest instruments. In particular, a new challenge that is emerging is how to fit previously solved high resolution protein structures into low resolution cryo-EM density maps. However, the problem here is that large molecular machines will have multiple sub-components, some of which will be unknown, and many of which will fit each part of the map almost equally well. Thus, the general problem of building high resolution 3D models from cryo-EM data is like building a complex 3D jigsaw puzzle in which several pieces may be unknown or missing, and none of which will fit perfectly. We wish to proceed firstly by putting more emphasis on the single-body terms in the scoring function 53, and secondly by using fast CG representations and knowledge-based distance restraints to prune large regions of the search space, as intiated with the EROS-DOCK software during the PhD project of Maria Elisa Ruiz Echartea (defended in 2019)13, 14.

3.2.6 Protein-Nucleic Acid Interactions

As well as playing an essential role in the translation of DNA into proteins, RNA molecules carry out many other essential biological functions in cells, often through their interactions with proteins. A critical challenge in modeling such interactions computationally is that the RNA is often highly flexible, especially in single-stranded (ssRNA) regions of its structure. These flexible regions are often very important because it is through their flexibility that the RNA can adjust its 3D conformation in order to bind to a protein surface. However, conventional protein-protein docking algorithms generally assume that the 3D structures to be docked are rigid, and so are not suitable for modeling protein-RNA interactions. There is therefore much interest in developing dedicated protein-RNA docking algorithms which can take RNA flexibility into account. This research topic has been initiated with the recruitement of Isaure Chauvot de Beauchêne in 2016 and is becoming a major activity in the team. A novel flexible docking algorithm is currently under development in the team. It first docks small fragments of ssRNA (typically three nucleotides at a time) onto a protein surface, and then combinatorially reassembles those fragments in order to recover a contiguous ssRNA structure on the protein surface 50, 51.

As the correctness of the initial docking of the fragments settles an upper limit to the correctness of the full model, we are now focusing on improving that step. A key component of our docking tool is the energy function of the protein-fragment interactions that is used both to drive the sampling (positioning of the fragments) by minimization, and to discriminate the correct final positions from decoys (i.e., false positives). We are developing a new approach to create knowledge-based parameters for coarse-grain energy functions from public structural data, in collaboration with Sjoerd de Vries (INSERM). Such approach will be applied first to ssRNA-protein complexes, then to other types of complexes such as protein-peptides.

Another key requirement for this approach is an exhaustive but non-redundant library of possible internal conformations of RNA fragments. Our library is built by clustering hundreds of thousands of experimentally known RNA structures, based on an approximate geometric similarity criteria. We have recently developped a new representation for the clustering of 3D conformations based on internal coordinates, in order to optimise the representativity of the library (internship of Alix Delannoy). We are currently developping a new clustering method based on epsilon-networks to reduce the cardinality of the ensemble of clusters.This is part of the PhD subject of Antoine Moniot, co-supervised by Yann Guermeur (ABC LORIA team).

In the future, we will improve the combinatorial algorithm used for reassembling the docked fragments using both experimental constraints (PhD project Anna Kravchenko) and machine-learning approaches (PhD project Hrishikesh Dondge).

7.1.1 HLA-3D-Diff

• Name:
  Interface for viewing and superimposing 3D structures of HLA antigens
• Keywords:
  Visualization, 3D structure, Proteins, Molecular dynamics, GUI (Graphical User Interface)
• Scientific Description:

  The HLA-3D-Diff software contains two parts: (1) The HLA-3D-Diff database: its goal is to provide easy access to data from multiple public sources and data generated by members of the project , (2) The HLA-3D-Diff visualisation interface: a web app for visualizing structures and molecular dynamics (MD) simulations of HLA molecules.

  The GitLab repository contains: - The source code and Docker environment for these tools , - Documentation for the database, including models, the data dictionary, and input data , - Documentation for the interface (a short video).

• Functional Description:
  This tool makes it possible to visually search for 3D structural differences that may explain incompatibility or unexpected compatibility between two distinct HLA antigens. It is differential because it makes it possible to superimpose two different 3D structures, and dynamic because it makes it possible to visualize trajectories simulated by molecular dynamics.
• URL:
  https://­gitlab.­inria.­fr/­capsid/­hla-3d-diff
• Contact:
  Marie-Dominique Devignes
• Participants:
  Louane Sigrist, Diego Amaya Ramirez, Marie-Dominique Devignes, Jean-Luc Taupin

8.1.1 Biomedical Knowledge Discovery

We updated and extended a preliminary study carried out by Zia Alborzi (Inria PhD student 2015-2018) during his PhD thesis and aimed at inferring protein domain interactions from multiples sources of protein-protein interactions 1, 34. Results have been deposited as open research data for reuse facility in the zenodo repository.

We continued our collaboration with clinicians at the CHRU Nancy in the frame of the RHU FIGHT-HF program. In particular, we contributed to a study aimed at clustering patients according to their echocardiographic measurements and at characterizing the obtained clusters with respect to heart failure propensity 23.

In the frame of the PraktikPharma ANR project coordinated by Adrien Coulet (Orpaileur team), we investigated adverse drug reaction mechanisms with knowledge graph mining 20. Finally we also participated in a study carried out in the Orpailleur team in the frame of the GeenAge IMPACT project to tackle the problem of identifying the most predictive and discriminant features in supervised classification problems 29.

8.1.2 Stochastic Random Trees for Similarity Computation

Kevin Dalleau defended his PhD thesis entitled “A stochastic approach based on random forests for dissimilarity calculation : application to clustering for various data structured” on 23 november 2021 52. The UET dissimilarity measure introduced in this thesis was used in a benchmarking study of clustering algorithms applied on mixed data (both categorical and numerical) 9.

8.1.3 Graph-based Approaches for Machine Learning and Protein Annotation

One of the methods developed by Bishnu Sarker for protein function annotation is based on label propagation in graphs. GrAPFI (Graph-based Automatic Protein Function annotation) relies on a protein graph in which edges are weighted by the domain similarity between proteins 76. When applied to function annotations taken from Gene Ontology, the method was improved by post-processing of the predicted annotations, leveraging the hierarchical structure of the ontology. Moreover it was shown that the post-processing step also improves other methods of protein function annotation 75.

A preliminary exploration study of the potential of generative adversarial networks (GAN) for protein function annotation was conducted by Bishnu Sarker. The proposed method “Prot-A-GAN” uses GAN-like adversarial training for learning embedding of nodes and relation in a heterogeneous knowledge graph 32. Prot-A-GAN trains a discriminator using domain-adaptive negative sampling to discriminate positive and negative triples, and then, it trains a generator to guide a random walk over the knowledge graph that identify paths between proteins and GO annotations.

All these contributions constitute the PhD Thesis of Bishnu Sarker whose defense took place on April 23, 2021 37.

In the context of the “TempoGraphs” project, we developed a similarity function for sparse binary data with application on protein function annotation 26. The proposed similarity function is based on the analysis of the best existing similarity functions for the protein function annotation task. We performed experiments in a simple pairwise similarity scenario and also using our proposal as part of a more complex protein function annotation method.

In the frame of Kamrul Islam's PhD project, we first tackled the problem of link prediction in large knowledge graphs. We conducted an experimental evaluation of similarity-based and embedding-based link prediction methods on graphs. The studied approaches were evaluated on several graph datasets with different properties from various domains. The precision of similarity-based approaches was computed in two different ways to highlight the difficulty of tuning the threshold for deciding the link existence based on the similarity score 22.

Then, we tackled the problem of negative sampling, in other words: how to sample negative triplets from non-observed ones in a training knowledge graph. We developed a simple negative sampling (SNS) method based on the assumption that the entities which are closer to the corrupted entity in the embedding space are able to provide high-quality negative triples. The proposed method has been evaluated through link prediction task on several popular knowledge graph datasets including a new biological knowledge graph dataset (FIGHT-HF-23R) 35.

Collaborative work on the development of a distributed and incremental algorithm for large-scale graph clustering has led to a technical report in 2020 38 which contributed to the completion of Wissem Inoubli's PhD thesis (defended early January 2021) 36.

Another collaborative work on the study of big services has led to a review paper in 2021 24. The purpose of the study is to provide an understanding of the new emerging big service model from the point of view of "lifecycle management phases". We also study the role of big data frameworks in the provisioning of big services 24.

8.1.4 Biological network modelling

An inter-disciplinary project, funded by the Université de Lorraine, is ongoing since 2018 in collaboration with the CRAN laboratory. It concerns the study of the role of estrogen receptors in the development of gliobastoma tumors. The available data is high-dimensional but involves rather small numbers of samples. The challenge is to identify relevant sets of genes which are differentially expressed in various phenotyped groups (w.r.t. gender, age, tumor grade). The objectives are to infer pathways involving these genes and to propose candidate models of tumor development which will be experimentally tested thanks to an ex-vivo experimental system available at the CRAN 21.

Boolean Networks (BNs) are a simple formalism used to study complex biological systems when the prediction of exact reaction times is not of interest. BNs play a key role in understanding the dynamics of the studied systems and in predicting their disruption in case of complex human diseases. BNs are generally built from experimental data and knowledge from the literature, either manually or with the aid of programs. The automatic synthesis of BNs is still a challenge for which several approaches have been proposed. We proposed ASKeD-BN, a new approach based on Answer-Set Programming to synthesise BNs constrained in their structure and dynamics 27. By applying ASKeD-BN on several well-known biological systems, we provide empirical evidence that our approach can construct BNs in line with the provided constraints. We compared our approach with three existing methods (REVEAL, Best-Fit and caspo-TS) and showed that our approach produces a small number of BNs which are covering a good proportion of the dynamical constraints, and that the variance of this coverage is low. The published paper on ASKed-BN was selected as a highlight oral presentation in the JOBIM 2021 conference 33.

Today SBML is the standard format to represent models of biological systems. Most of the established curated models available in the Biomodels repository are quantitative, but in some cases qualitative models, such as boolean networks are preferable. We proposed to focus on the automatic transformation of quantitative SBML models to boolean networks through the so-called SBML2BN pipeline. By running SBML2BN on more than 200 quantitative SBML models, we provided evidence that we can automatically construct Boolean networks which are compatible with both the structure and the dynamics of a given quantitative SBML model 28, 34.

8.2.1 Clustering of 3D conformations

Our fragment-based approach to dock RNA on proteins requires libraries of 3D conformations of RNA fragments. These must be representative of the full ensemble of possible 3D conformations of RNA fragments but be of small enough cardinality to be usable for combinatorial assemblies of fragments into RNAs. Therefore, they must approximate all possible local RNA conformations within a given precision, by a set of well chosen representative fragments. Such a set can be obtained by clustering a larger set of fragments (typically ${10}^4$ to ${10}^5$) whose structures have been solved experimentally, using suitable clustering algorithm and measure of dissimilarity between fragments. A commonly used measure of dissimilarity in structural biology is the root mean square deviation (RMSD), whose exact computation requires a pairwise structural alignment. But this alignment is highly time-consuming and not applicable for a very large initial set of fragments. We have developed an approach based on feature extraction to perform an effective clustering, while avoiding a computationally expensive full pairwise alignment. Using as example poly-A RNA fragments of 3 nucleotides (3-nt), we searched for internal coordinates whose differences can best approximate the RMSD between two fragments without any superposition. We found that the simple differences of internal distances and angles can provide a lower bound on the RMSD, allowing us to filter out redundant fragment pairs without any RMSD computing 30. This was developed by Alix Delannoy (3A Mines research training), co-supervised by Antoine Moniot, Isaure Chauvot de Beauchêne and Yann Guermeur. Concerning the clustering step, we are currently working on an algorithm to compute epsilon-nets of small cardinality for finite sets of points. Each of these points should be at a distance inferior to epsilon to a point in the net. The minimal cardinality that can be achieved is called the (epsilon) covering number. Although the literature provides us with upper bounds on the covering numbers, the so-called extended Sauer-Shelah lemmas, the proofs are non constructive. To the best of our knowledge, there are no (efficient) algorithms to compute epsilon-nets of small cardinality in the framework of interest. We have chosen to develop an algorithm, based on two ideas: the ascending hierarchical clustering (AHC) and the minimal enclosing spheres. The hierarchical approach makes it possible to go from a feasible solution to another and stop as soon as no more fusion is possible 41, 42. This work is carried out by Antoine Moniot for his PhD thesis, to be defended in 2022.

8.2.2 Modeling and design of RNA-protein complexes

Our H2020 ITN project RNAct aims at designing new RNA-binding proteins based on the well-conserved protein domain called RNA Recognition Motif (RRM). In this frame, we have created in 2020 the InteR3Mdb database that contains all known 3D structures of RRM and RRM-RNA complexes. We have now superimposed all those structures and aligned their sequences (which is not trivial due to sequence and structure variability, large deletions, inconsistencies in the raw data...), and identified all nucleotide – amino acid contacts in each experimental structure of an RRM-RNA complex. For each contact, the amino acid type and position in the sequence, the nucleotide type and position in the sequence, the type of interaction (hydrogen bond, ionic bond, stacking...), the atoms involved and the atom-atom distances were identified. This information is extremely useful and provides an important resource on its own to the scientific community. Importantly, it allowed us to infer a contact propensity for each amino acid at every conserved position in the sequence, and toward a given nucleotide type (A/C/U/G). This is currently used to computationally design and propose point mutations to alter RNA binding specificity. Synthesis of RRM proteins carrying such mutations is underway within the RNAct Consortium in view of performing binding assays that can validate the predictions.

In parallel, we exploited the InteR3Mdb database to tackle the bottleneck encoutered when modeling RNA-protein complexes using the known 3D shape of the unbound protein. Indeed, the changes that occur in the protein 3D shape when it binds RNA are hard to predict. In InteR3Mdb (Hrishikesh Dhondge's PhD thesis), the protein-bound or protein-unbound status of all known RRM 3D structures have been added as new attributes for each database entry. Then, data was processed to infer those structural characteristics that could distinguish bound from unbound RRMs. Results can be used for the prediction of the bound conformation of an RRM from its unbound conformation. The knowledge and prediction power gained from the bound/unbound comparison of RRMs will enable us to (i) unravel structural determinants of the RNA binding capacities of RRMs, (ii) address the main limitation of RRM – RNA docking approaches, and (iii) take into account conformational changes associated with RNA binding in the design of RRMs with altered RNA binding specificity.

In collaboration with Sjoerd de Vries (INSERM), we have implemented a new method to create energy parameters for RNA-protein interactions in coarse-grained representations. Each amino-acid of the protein and each nucleotide of the RNA is represented by 2 to 7 pseudo-atoms (“beads”). For each model of an RNA - protein interaction, the energy is computed as the sum of the bead-bead energies, and the model with the lowest energy is considered as the most probable. Each bead-bead energy depends solely on the 2 bead types (among 17 RNA types and 32 protein types) and the inter-bead distance. For each pair of bead types, the energy function has the shape of a Lennard-Jones potential, given by 2 parameters that determine the distance of minimal energy and the value of the minimal energy. The current parameters were extracted in 2010 by statistics on existing RNA-protein crystal structures and optimized by a random Monte Carlo-like strategy. These parameters were not initially tailored to single-stranded RNA and their performance is not flawless. A main goal of Anna Kravchenko's thesis is to optimize those parameters for ssRNA. To achieve that, we set up a novel histogram-based approach. For each pair of bead types, we (a) convert the current energy function into a log-odds histogram of the expected occurrences of distances (discretized into bins) in correct/incorrect models (from an in-house benchmark), using the Boltzmann equation; (b) obtain the corresponding histogram on a benchmark-wide docking test, which corresponds to the residual error of the energy function; (c) sum the predicted and real histograms; (d) analytically fit the energy parameters to the resulting histogram; repeat until convergence, i.e. until the residual histogram is flat. A first prototype of this workflow has been validated on a small training set 40, and is now being extended to our full benchmark of hundred of protein-RNA complexes.

8.2.3 3D Modeling and Virtual Screening

3D modeling has been revolutionized by the AlphaFold2 program created by DeepMind using deep learning methods. We used the open access AlphaFold2 program to confirme and refine our models of the Relaxase RelSt3 dimer in the frame of the CITRAM project. The resulting model was used to simulate the interaction with both ss and ds RNA. This work allowed Dominique Mias-Lucquin to win the best Poster award at GGMM'2021 (An homodimeric multi-domain protein interacting with ssDNA and dsDNA: The challenges of RelSt3 relaxase modelisation).

Virtual screening of small molecules libraries is an essential part of the team's expertise that is often called upon by external partners in biology. Our work on the search of inhibitors of beta1,4 galactosyltransferase 7 was presented at the RARE meeting 39.

10.1.1 Inria associate team not involved in an IIL or an international program

10.1.2 Participation in other International Programs

10.2.1 FP7 & H2020 projects

3D structural differences among HLA proteins

Participants: Marie-Dominique Devignes [contact person], Diego Amaya Ramirez, Louane Sigrist, Bernard Maigret.

Context : PhD thesis Inria-Inserm of Diego Amaya Ramirez (2019-2022). This thesis (HLA genetic system and organ transplantation: understanding the basics of immunogenicity to improve donor - receptor compatibility when assigning grafts to recipients) is co-supervised by Marie-Dominique Devignes and Jean-Luc Taupin (Professor of Immunology at the Laboratory of Immunology and Histocompatibility at Saint-Louis Hospital and Diderot University in Paris).

Predicting the 2D structure of protein-bound RNAs

Participants: Isaure Chauvot de Beauchêne.

Context: Co-supervision of a M2 internship with Fariza Tahi and Eric Angel (AROBAS team) at the IBISC lab (University of Evry). This project consisted in integrating the CAPSID library of protein-bound RNA fragments into the 2D RNA prediction pipeline of the AROBAS team to create specific RNA predictions in the protein-binding context. It provided a proof-of-principle, and a demand for a PhD grant has been submitted to pursue the project in 2022.

FIGHT-HF

Participants: Marie-Dominique Devignes [contact person], Malika Smaïl-Tabbone [contact person], Emmanuel Bresso, Bernard Maigret, Sabeur Aridhi, Kévin Dalleau, Claire Lacomblez, Gabin Personeni, Philippe Noel.

Project title: Combattre l’insuffisance cardiaque : Projet de Recherche Hospitalo-Universitaire FIGHT-HF; PI: Patrick Rossignol, Université de Lorraine (FHU-Cartage); Value: 9 M€ (CAPSID and Orpailleur: 450 k€, approx); Duration: 2015–2021. Description: This “Investissements d'Avenir” project aims to discover novel mechanisms for heart failure and to propose decision support for precision medicine. The project has been granted 9 M€, and involves many participants from Nancy University Hospital Federation “CARTAGE”. Marie-Dominique Devignes and Malika Smaïl-Tabbone are coordinating a work-package dedicated to network-based science, decision support and drug discovery for this project.

IFB

Participants: Marie-Dominique Devignes [contact person], Sabeur Aridhi, Isaure Chauvot de Beauchêne.

Project title: Institut Français de Bioinformatique; PI: Claudine Médigue and Jacques van Helden (CNRS UMS 3601); Value: 20 M€ (CAPSID: 126 k€); Duration: 2014–2021. Description: The CAPSID team is associated with the IFB (Institut Français de Bioinformatique), the French national network of bioinformatics platforms. The principal aim is to make bioinformatics skills and resources more accessible to French biology laboratories. Marie-Dominique Devignes is coordinating with Alban Gaignard the Interoperability task in the Integrative Bioinformatics Workpackage.

Anti-covid research

Participants: Bernard Maigret [contact person], Marie-Dominique Devignes, Dominique Mias-Lucquin, Isaure Chauvot de Beauchêne.

Our research on the structural molecular determinants of binding of SARS-Cov-2 variant Spike proteins to their cellular receptor has lead us to develop partnerhsips with selected teams in France.

• LAMA : Laboratoire de Mathématique, CNRS — Université Savoie Mont Blanc UMR 5127. Persons involved : Laurent Vuillon, Aria Gheeraert.
• IRIM : Institut de Recherche en Infectiologie de Montpellier, CNRS-Université de Montpellier UMR 9004. Persons involved : Laurent Chaloin, Olivier Montcorgé.

CPER IT2MP Santé (2016-2021)

In the regional Contrat de Plan Etat-Région Santé (Innovations Technologiques, Modélisation et Médecine Personnalisée), MD Devignes and Malika Smaïl-Tabbone coordinate the SMEC platform (Simulation, Modélisation, Extraction de Connaissances). This has strenghtened our partnership with two teams at the regional level.

• LPCT (Laboratoire de Physique et Chimie Théoriques, CNRS-Université de Lorraine UMR 7019), Team of Chris Chipot and François Dehez.
• CIC-P (Centre d'Investigation Clinique Plurithématique, CHRU de Nancy) and DCAC (Défailance Cardiovasculaire Aiguë et Chronique, Inserm - Université de Lorraine U1116), Patrick Rossignol, Nicolas Girerd, Patrick Lacolley and Véronique Régnault.

LUE IMPACT projects (2017-2021)

The CAPSID team has been associated with two LUE projects that terminated in 2021

• IMPACT GeenAge. MD Devignes (contact person). Partners involved : IMOPA (Ingénierie Moléculaire et Physiopathologie Articulaire, CNRS - Université de Lorraine, UMR 7365), Isabelle Behm ; Orpailleur (CRI Nancy Grand-Est), Amedeo Napoli, Alexandre Bazin (GeenAge post-doc).
• IMPACT BioMolecules. Sabeur Aridhi (contact person). Partners involved : Laboratoire d'Ingenierie des Biomolecules (ENSAIA - Université de Lorraine), Freédéric Borges ; Orpailleur (CRI Nancy Grand-Est), Yannick Toussaint.

LUE-FEDER CITRAM (2017-2022)

The CITRAM project (Conception d’Inhibiteurs de la Transmission des Résistances Anti-Microbiennes) was created as a LUE project in 2017 and later extended with FEDER co-funding. It involves two other teams:

• DynAMIc lab (Genome dynamics and microbial adaptation, INRAE - Université de Lorraine UMR 1128), Team of Nathalie Leblond and Nicolas Soler.
• LPCT (Laboratoire de Physique et Chimie Théoriques, CNRS-Université de Lorraine UMR 7019), Team of Chris Chipot and François Dehez.

11.1.1 Scientific events: organisation

11.1.2 Scientific events: selection

11.1.3 Journal

11.1.4 Leadership within the scientific community

• Marie-Dominique Devignes is co-responsible of the Interoperability task force at the French Institute of Bioinformatics and participant of the ELIXIR Interoperability platform.
• Marie-Dominique Devignes proposed and co-organized a mini-symposium on Open Science and Interoperability at the JOBIM conference, 6-9 July 2021, Paris.
• Isaure Chauvot de Beauchêne co-organized a mini-symposium on "La modélisation structurale intégrative à l’ère des mégadonnées et de l’intelligence artificielle" at the JOBIM conference, 6-9 July 2021, Paris.
• Malika Smaïl-Tabbone was invited to chair a session at the 10th International Conference on Complex Networks and their Applications, November 30 - December 2, 2021, Madrid.

11.1.5 Scientific expertise

Marie-Dominique Devignes reviewed applications grants for a call from Nouvelle Aquitaine Region.

11.2.1 Teaching

• Malika Smaïl-Tabbone is associate professor at the Université de Lorraine with a full service. She is co-responsible with Pascal Moyal of the IMSD track ("Ingénierie Mathématique pour la Science des Données") in the Applied Mathematics Master's degree at the Université de Lorraine. She is also member of the pedagogic team of the CMI BSE ("Cursus Master Ingénieur Biologie-Santé-Environnement") and also in charge of corporate relations.
• Sabeur Aridhi is associate professor at the Université de Lorraine with a full service. He is responsible for the major in IAMD ("Ingénierie et Applications des Masses de Données") at TELECOM Nancy.
• Marie-Dominique Devignes teaches every year 10 to 16h in the CMI BSE.
• Athenaïs Vaginay started an ATER position in september 2021.

11.2.2 Supervision

• Isaure Chauvot de Beauchêne supervised the 3A Mines internship of Alix Delannoy (co-supervised with Antoine Moniot and Yann Guermeur) and the M2 internship of Nathalie Bernard (co-supervised with Fariza Tahi and Eric Angel at IBISC, Evry). She also supervised the secondment of Anna Perez Rafols in the frame of the ITN RNAct.
• Marie-Dominique Devignes co-supervised with Isaure Chauvot de Beauchêne the internship of Karina Pats (exchange ITMO St-Petersbourg - Ecole des Mines Nancy) and with Anne Gégout-Petit (IECL) the M2 internship of Dalil Rouabah. She also supervised the MIASHS L3 internship of Valentin Retter and the CMI BSE L3 internship of Nathan Nourdin.
• Sabeur Aridhi supervised the M2 internship of Amal Stiti.
• Malika Smaïl-Tabbone co-supervised with Hélène Dumond the M2 internship of Nolwenn Lebourdais (affected to CRAN).
• Bernard Maigret supervised the M1 internship (initiation to research) of Nicolas Bombarde and Ugo Cottin.

11.2.3 Juries

• Marie-Dominique Devignes: reviewer in the PhD defense committee of Julie Lao, Université de Lorraine, 23 Feb 2021, examinator in the PhD defense committee of Louis Becquey, University of Evry, 6 Oct 2021, and reviewer in the PhD defense committee of Floris Chabrun, University of Angers, 7 Dec 2021.
• Marie-Dominique Devignes and Sabeur Aridhi: co-supervisors in the PhD defense committee of Bishnu Sarker, Université de Lorraine, 23 April 2021.
• Sabeur Aridhi: Co-supervisor in the PhD defense committee of Wissem Inoubli, University of Tunis, 7 Jan 2021.
• Malika Smaïl-Tabbone: invited member in the PhD defense committee of Nicolas Scalzitti, University of Strasbourg, 29 Sept 2021, co-supervisor in the PhD defense committee of Kevin Dalleau, Université de Lorraine, 23 Nov 2021.

11.3.1 Internal or external Inria responsibilities

• Isaure Chauvot de Beauchêne was member of the hiring committee of the Inria competition for researcher recruitment.
• Isaure Chauvot de Beauchêne was member of two selection committees for associate professor positions at Université de Reims Champagne-Ardenne and Université de Lyon.

International journals

• 18 articleA. A.Ahoefa Ablavi AWUSSI, E.Emeline Roux, C.Catherine Humeau, Z.Zeeshan HAFEEZ, B.Bernard Maigret, O. K.Oun Ki Chang, X.Xavier Lecomte, G.Gérard Humbert, L.Laurent Miclo, M.Magali Genay, C.Clarisse Perrin and A.Annie Dary-Mourot. Role of the Sortase A in the Release of Cell-Wall Proteinase PrtS in the Growth Medium of Streptococcus thermophilus 4F44.Microorganisms911November 2021, 2380
  HALDOI
• 19 articleS. Z.Seyed Ziaeddin Alborzi, A.Amina Ahmed Nacer, H.Hiba Najjar, D.David Ritchie and M.-D.Marie-Dominique Devignes. PPIDomainMiner: Inferring domain-domain interactions from multiple sources of proteinprotein interactions.PLoS Computational Biology1782021, e1008844
  HALDOI
• 20 articleE.Emmanuel Bresso, P.Pierre Monnin, C.Cédric Bousquet, F.-E.François-Elie Calvier, N.-C.Ndeye-Coumba Ndiaye, N.Nadine Petitpain, M.Malika Smaïl-Tabbone and A.Adrien Coulet. Investigating ADR mechanisms with Explainable AI: a feasibility study with knowledge graph mining.BMC Medical Informatics and Decision Making211May 2021, 171
  HALDOIback to text
• 21 articleA.Alex Hirtz, N.Nolwenn Lebourdais, F.Fabien Rech, Y.Yann Bailly, A.Athénaïs Vaginay, M.Malika Smaïl-Tabbone, H.Hélène Dubois-Pot-Schneider and H.Hélène Dumond. GPER agonist G-1 disrupts tubulin dynamics and potentiates temozolomide to impair glioblastoma cell proliferation.Cells10December 2021, 3438
  HALDOIback to text
• 22 articleM. K.Md Kamrul Islam, S.Sabeur Aridhi and M.Malika Smaïl-Tabbone. An Experimental Evaluation of Similarity-Based and Embedding-Based Link Prediction Methods on Graphs.International Journal of Data Mining & Knowledge Management Process11September 2021, 1 - 18
  HALDOIback to text
• 23 articleM.Masatake Kobayashi, O.Olivier Huttin, M.Martin Magnusson, J. P.João Pedro Ferreira, E.Erwan Bozec, A.-C.Anne-Cecile Huby, G.Gregoire Preud'homme, K.Kevin Duarte, Z.Zohra Lamiral, K.Kevin Dalleau, E.Emmanuel Bresso, M.Malika Smaïl-Tabbone, M.-D.Marie-Dominique Devignes, P. M.Peter M Nilsson, M.Margret Leosdottir, J.-M.Jean-Marc Boivin, F.Faiez Zannad, P.Patrick Rossignol and N.Nicolas Girerd. Machine Learning-Derived Echocardiographic Phenotypes Predict Heart Failure Incidence in Asymptomatic Individuals.JACC: Cardiovascular ImagingS1936-878X21September 2021, 00556-8
  HALDOIback to text
• 24 articleH.Haithem Mezni, M.Mokhtar Sellami, S.Sabeur Aridhi and F. B.Faouzi Ben Charrada. Towards big services: a synergy between service computing and parallel programming.Computing10311November 2021, 2479-2519
  HALDOIback to textback to text
• 25 articleG.Gregoire Preud'homme, K.Kévin Duarte, K.Kevin Dalleau, C.Claire Lacomblez, E.Emmanuel Bresso, M.Malika Smaïl-Tabbone, M.Miguel Couceiro, M.-D.Marie-Dominique Devignes, M.Masatake Kobayashi, O.Olivier Huttin, J. P.João Pedro Ferreira, F.Faiez Zannad, P.Patrick Rossignol and N.Nicolas Girerd. Head-to-head comparison of clustering methods for heterogeneous data: a simulation-driven benchmark.Scientific Reports111February 2021, 4202
  HALDOI
• 26 articleM. B.Marcelo B.A. Veras, B.Bishnu Sarker, S.Sabeur Aridhi, J. P.João P.P. Gomes, J. A.José A.F. Macêdo, E. M.Engelbert Mephu Nguifo, M.-D.Marie-Dominique Devignes and M.Malika Smaïl-Tabbone. On the design of a similarity function for sparse binary data with application on protein function annotation.Knowledge-Based Systems238February 2022, 107863
  HALDOIback to text

International peer-reviewed conferences

• 27 inproceedingsA.Athénaïs Vaginay, T.Taha Boukhobza and M.Malika Smaïl-Tabbone. Automatic synthesis of boolean networks from biological knowledge and data.author's manuscript of OLA 2021, CCIS 1443 proceedingsOLA 2021 - International Conference of Optimization and LearningCatane, ItalyJune 2021
  HALback to text
• 28 inproceedingsA.Athénaïs Vaginay, T.Taha Boukhobza and M.Malika Smaïl-Tabbone. From quantitative SBML models to boolean networks.author's manuscript of Studies in Computational Intelligence 1016, SpringerComplex Networks & Their Applications XVolume 2, Proceedings of the TenthInternational Conference on ComplexNetworks and Their ApplicationsCOMPLEX NETWORKS 2021----------Rosa Maria Benito ·Chantal Cherifi · Hocine Cherifi ·Esteban Moro · Luis M. Rocha ·Marta Sales-Pardo EditorsCNA 2021 - 10th International Conference on Complex Networks and their ApplicationsMadrid, SpainNovember 2021
  HALback to text

Conferences without proceedings

• 29 inproceedingsA.Alexandre Bazin, M.Miguel Couceiro, M.-D.Marie-Dominique Devignes and A.Amedeo Napoli. A Hybrid Approach to Identifying the Most Predictive and Discriminant Features in Supervised Classification Problems.ICCS 2021 - 26th International Conference on Conceptual StructuresVirtual, FranceSeptember 2021
  HALDOIback to text
• 30 inproceedingsA.Alix Delannoy, A.Antoine Moniot, Y.Yann Guermeur and I.Isaure Chauvot de Beauchêne. Feature extraction for the clustering of small 3D structures: application to RNA fragments.JOBIM 2021 - Journées Ouvertes en Biologie, Informatique et MathématiquesParis/Virtual, FranceJuly 2021, 1-6
  HALback to text
• 31 inproceedingsM. K.Md Kamrul Islam, S.Sabeur Aridhi and M.Malika Smaïl-Tabbone. Appraisal Study of Similarity-Based and Embedding-Based Link Prediction Methods on Graphs.CDKP 2021 - 10th International Conference on Data Mining & Knowledge Management ProcessLondon, United KingdomAIRCC Publishing CorporationJuly 2021, 81-92
  HALDOI
• 32 inproceedingsB.Bishnu Sarker, M.-D.Marie-Dominique Devignes, G.Guy Wolf and S.Sabeur Aridhi. Prot-A-GAN: Automatic Protein Function Annotation using GAN-inspired Knowledge Graph Embedding.ICML 2021 - Workshop on Computational BiologyVirtual, United StatesJuly 2021
  HALback to text
• 33 inproceedingsA.Athénaïs Vaginay, T.Taha Boukhobza and M.Malika Smaïl-Tabbone. Automatic synthesis of boolean networks from biological knowledge and data.JOBIM 2021 - Journées Ouvertes en Biologie, Informatique et MathématiquesParis, FranceJuly 2021
  HALback to text
• 34 inproceedingsA.Athénaïs Vaginay, T.Taha Boukhobza and M.Malika Smaïl-Tabbone. From quantitative SBML to boolean networks.CMSB 2021 - 19th conference on Computational Methods in Systems Biology,Bordeaux, FranceSeptember 2021
  HALback to textback to text

Scientific book chapters

• 35 inbookM. K.Md Kamrul Islam, S.Sabeur Aridhi and M.Malika Smaïl-Tabbone. Simple negative sampling for link prediction in knowledge graphs.1016Complex Networks & Their Applications XStudies in Computational IntelligenceSpringer International PublishingJanuary 2022, 549-562
  HALDOIback to text

Doctoral dissertations and habilitation theses

• 36 thesisW.Wissem Inoubli. Analysis and Mining of Large Dynamic Graphs: case of graph clustering.Université de Tunis El Manar (Tunisie)January 2021
  HALback to text
• 37 thesisB.Bishnu Sarker. On Graph-Based Approaches for Protein Function Annotation and Knowledge Discovery.Université de LorraineApril 2021
  HALback to textback to text

Reports & preprints

• 38 miscW.Wissem Inoubli, S.Sabeur Aridhi, H. ·.Haithem · Mezni, M.Mondher Maddouri and E.Engelbert Mephu Nguifo. A Distributed and Incremental Algorithm for Large-Scale Graph Clustering.2021
  HALback to text

Other scientific publications

• 39 inproceedingsS.Sandrine Gulberti, C.Christel Valencia-Schmitt, P.Pascal Villa, I.Isaure Chauvot de Beauchêne, B.Bernard Maigret and S.Sylvie Fournel-Gigleux. Vers un nouveau traitement des mucopolysaccharidoses ? Recherche d’inhibiteurs de la β1,4-galactosyltransférase 7 (β4GalT7) par des approches combinées de criblage expérimental et virtuel.Rencontres RARE 2021Paris/virtuel, FranceOctober 2021
  HALback to text
• 40 inproceedingsA.Anna Kravchenko, M.Malika Smaïl-Tabbone, I.Isaure Chauvot de Beauchêne and S. J.Sjoerd Jacob De Vries. New strategy for optimizing knowledge-based docking parameters: application to ssRNA-protein docking.JOBIM 2021 - Journées Ouvertes en Biologie, Informatique et MathématiquesParis, FranceJuly 2021
  HALback to text
• 41 inproceedingsA.Antoine Moniot, I.Isaure Chauvot de Beauchêne and Y.Yann Guermeur. Agglomerative clustering of fragment 3D structures based on pairwise RMSD.ISMB ECCB 2021Virtual, FranceJuly 2021
  HALback to text
• 42 inproceedingsA.Antoine Moniot, I.Isaure Chauvot de Beauchêne and Y.Yann Guermeur. New clustering method to infer prototypes covering the 3D structures of nucleic acid fragments.RECOMB 2021 - 25th International conference on research in computational molecular biologyPadova/Virtual, ItalyAugust 2021
  HALback to text