Description du sujet de thèse

The Camille Jordan Institute (ICJ, UMR CNRS 5208, CNRS, Claude Bernard University Lyon 1) is offering a full-time PhD position on the topic of “Topological analysis of three-dimensional protein structures and topological neural networks for the analysis of organisms' life history traits.”

This PhD thesis aims to develop methods at the interface of molecular biotopology and machine learning to identify the molecular and structural determinants of piezoresistance.

Molecular biotopology involves studying biological macromolecules, mainly proteins, and their functional properties using techniques derived from topological data analysis (TDA) and spectral geometry. In this thesis, we will propose new multi-scale representations of these macromolecules and determine their geometric (curvature markers) and spectral (spectral markers) characteristics. These markers, constructed from discrete curvatures (Forman-Ricci, Ollivier-Ricci), the persistent Laplacian, and structural, genomic, and environmental factors, constitute vector representations of macromolecules.

Appropriate metrics will be defined to train predictive models on these representations, combining deep learning and statistical regression, to better understand the mechanisms of piezoresistance.

The thesis will also focus on modeling the evolution of these markers, taking into account the life history traits of organisms, using methods combining molecular phylogeny and ancestral sequence reconstruction. The aim is to trace the joint evolution of structures, protein sequences and topological markers along phylogenetic trees, in order to test the proposed models.

This thesis thus combines mathematical modeling, evolutionary biology and machine learning to explore the mechanisms of adaptation to high pressure.
Expected results include:
- The development of learning models capable of predicting life-history traits from protein structure;
- Identification of molecular signatures associated with piezoresistance;
- Integration of theoretical and software advances into the DeltaFold environment.

Contexte de travail

The thesis project is part of an interdisciplinary approach to modeling and predicting the life-history traits of organisms based on the three-dimensional structures of proteins. It is financed as part of a project of the Mission pour les Initiatives Transverses et Interdisciplinaires (MITI) of the CNRS.

The thesis will be carried out within a multidisciplinary team on the LyonTech-la Doua science campus in Lyon. This team brings together mathematicians and biologists from three laboratories: the Institut Camille Jordan (ICJ, UMR CNRS 5208, CNRS, Université Claude Bernard Lyon 1), the Laboratoire de Biométrie et Biologie Évolutive (LBBE, UMR CNRS 5558, CNRS, Université Claude Bernard Lyon 1, VetAgro Sup), and the Laboratoire de Microbiologie, Adaptation et Pathogénie (MAP, UMR CNRS 5240, CNRS, Université Claude Bernard Lyon 1, INSA Lyon).

The thesis will be part of the DeltaFold project, which aims to develop new mathematical, algorithmic and software tools combining topology, machine learning and molecular biology, in order to identify the structural signatures of proteins associated with life-history traits.

The methods developed as part of the thesis will be implemented in the DeltaFold library, developed in Python by an interdisciplinary team from ICJ and LBBE. This library provides a development environment for programming predictive models in evolutionary biology, based on vectorizations of three-dimensional protein structures, via persistent homology.

The piezometric predictive models will be validated by comparison with experimental results produced within the MAP laboratory. These validations will be based on two prokaryotic models: Thermococcales (81 species, each corresponding to around 2,000 proteins, i.e. a dataset of around 162,000 proteins) and Alteromonadales (1,000 species, each corresponding to around 6,000 proteins), for which a realistic subset centered on the genera presenting piezophilic organisms, Shewanella and Colwellia, will be selected.

Contraintes et risques

We are looking for a candidate (M/F) with a solid academic record and a Master's degree in mathematics or bioinformatics. An excellent theoretical background is essential. Previous experience in topological data analysis (TDA), statistics or data science will be a welcome asset.

Skills / Qualifications.
- Master's degree or equivalent in mathematics or bioinformatics.
- Good command of written and spoken English.
- Organizational skills and ability to engage in interdisciplinary and collaborative work.
- Willingness to participate in seminars, workshops and scientific meetings.

Candidates will be integrated into a multidisciplinary team. The candidate (M/F) must therefore possess excellent personal skills and be able to work in a team.

Pre-selection will be based on CV, experience, skills and cover letter. Shortlisted candidates will be invited to an audition before a selection committee.

Applications must include the following documents:
- A complete Curriculum Vitae,
- A letter of motivation (1 page), setting out the candidate's research interests in relation to the project,
- One or two letters of recommendation,
- Transcripts of Master's degrees (M1 and M2) or equivalent, and any relevant certificates.