RubisCO.2 - Engineering new RubisCOs with increased CO2 catalytic activity

PRISM programme

The PRISM (PhD Research Programme for International Training in Sustainable Soft Matter) programme has launched its first call for applications, offering up to 14 fully funded PhD fellowships starting from 1 March 2027 at Paris Sciences & Lettres (PSL) University. The programme trains researchers to address ecological transition challenges through sustainable soft matter science, with projects focused on eco-friendly chemical processes, circular economy, renewable energies, and carbon capture, storage, and valorisation. Co-funded by the European Union under Horizon Europe MSCA COFUND (Grant Agreement 101261637) and partner institutions, PRISM provides interdisciplinary, international, and intersectoral training, including mobility opportunities, secondments, and courses in sustainability, innovation, entrepreneurship, career development, and transferable skills.

Applications must be submitted only via the PRISM website (https://prism.psl.eu/en/) between 1st September to 31 October 2026 (23:59 Paris time).                                                                    

The PhD project

Context: Photosynthetic carbon fixation is limited by the inefficiency of the RubisCO protein, at the core of the Calvin-Benson cycle. Empirical attempts to enhance efficiency by modifying current RubisCO’s through mutations have not been successful so far. The RubisCO.2 project aims to avoid the strong constraints in current RubisCO proteins by considering putative ancestral proteins, thought to be much more flexible [Schulz et al., Science (2022)]. New and efficient RubisCO’s will then be designed, using statistical-physics modeling, generative AI and in vivo directed evolution. This PhD will take place at LPENS in the Cocco-Monasson team, with a track record on the design of protein [Russ et al. (2020); Malbranke et al. (2023)] and enzymatic RNA [Fernandez-de-Cossio-Diaz et al. (2025)], in close collaboration with partners in biochemistry and biology labs in Sorbonne Université and Institut Pasteur. This strongly interdisciplinary consortium combines the expertise required for such a challenging project, and offers great opportunities to the DC for adapting the project across the 3-year duration of the PhD.

Objectives: The DC will develop predictive models to navigate RubisCO’s sequence-function landscape and improve its catalytic performance. The models will be both physics-grounded and data-driven, combining bio-physical/chemical information, and biological sequence and structure data to capture large-scale evolutionary constraints and propose novel, functional variants of RubisCO. Models will also integrate experimental datasets from algal evolution, which estimate mutational effects on stability and catalytic efficiency. Using these models, the DC will infer ancestral protein sequences [Thornton et al. (2004)], and then identify mutation combinations that will bypass evolutionary bottlenecks and allow for changing the balance between CO₂/O₂ specificity and catalytic speed.

Connections with experiments: The DC will implement interpretable AI models to propose de novo RubisCO designs, prioritizing variants for experimental validation. Collaboration with Pierre Crozet (CQSB, SU and Paris Biofoundry) [Crozet et al. (2018)] and David Bikard (Synthetic Biology, Institut Pasteur) [Rochette et al. (2026)] will allow the DC to test these AI-designed variants in vivo in the Chlamydomonas algae, a model organism for photosynthesis, and to use the results to retrain models and improve predictive accuracy. A secondment in R. Ranganathan’s lab in Chicago will enable the student to incorporate advanced methods for epistasis mapping and evolutionary landscape analysis, enriching the PhD’s computational toolkit. The connections with the experimental partners, all involved in startup creation, will be instrumental in exposing the DC to valorization environments.

Expected outcome: This PhD will bridge Physics-Informed AI models and experiments at the interface of physics, biochemistry, and engineering. The project will produce predictive frameworks for RubisCO performance, linking sequence to function, as well as novel RubisCO variants with improved catalytic properties, validated experimentally. We expect the acquired expertise on AI-guided enzyme design to be very valuable on the job market after completion of the PhD, either in an industrial or academic context.

3i dimensions

INTERNATIONAL: The RubisCO.2 PhD project embodies a strong international dimension through its planned 3-month secondment of the DC to R. Ranganathan’s laboratory at the University of Chicago, a renowned expert in protein evolution and computational biology and a long-standing collaborator of the Cocco-Monasson group at LPENS [Russ et al. (2020)]. This collaboration leverages Ranganathan’s pioneering work on epistasis, protein sequence-function relationships, and evolutionary landscapes—directly aligning with the project’s AI-driven protein design goals. The secondment will enable the student to integrate the computational approaches (statistical modeling of mutational effects) developed at LPENS and computational/experimental validation techniques developed in Ranganathan’s lab, enriching the PhD’s focus on predictive modeling for Rubisco optimization. This international exposure will broaden the student’s expertise, enhance the project’s innovative potential, and strengthen global collaborations in computational biology and synthetic evolution.

INTERSECTORAL: RubisCO.2 integrates a strong intersectoral dimension through the collaborations with the groups at Institut Pasteur and in the Paris Biofoundry, a cutting-edge platform bridging academia and industry. P. Crozet and D. Bikard have proven track records in innovation/valorization as they founded Biomemory (DNA data storage), Neoplants (engineered plants for air quality), and Eligo Bioscience (gene-editing therapies). These startups demonstrate the consortium’s ability to translate fundamental research into applied biotechnology. In addition, the Cocco-Monasson team supervising the DC is also taking part to the PariSanté project, in which PSL is a key actor, with strong connections with the biomedical industry in the Paris area. Several alumni have already joined startups (e.g. Phagos) or large AI companies (Meta, Google). The project’s focus also strongly aligns with industrial interests in sustainable bioproduction and carbon fixation, opening avenues for partnerships with agritech or synthetic biology companies.

INTERDISCIPLINARY: The RubisCO.2 project is highly interdisciplinary, seamlessly integrating physics, biochemistry, synthetic biology, and artificial intelligence to address a fundamental challenge in carbon fixation and photosynthesis. The project relies on physics-based tools and concepts (statistical physics approaches, both analytical [Mauri et al. (2023)] and numerical [Huot et al. (2026)]) and AI-driven design (generative models) to predicts and optimizes mutations capable of exploring Rubisco’s sequence-function landscape, in close connection with experiments done by biologists (directed evolution in Chlamydomonas, in vivo diversification with hypermutating systems). Structural biology (crystallography, synchrotron access) further bridges chemistry and biology to validate enzyme improvements. This convergence allows the project to transcend traditional boundaries: physics informs AI models, biology provides experimental data, and synthetic biology implements innovations. The iterative design-build-test-learn cycle exemplifies this synergy, where computational predictions guide experiments, and biological insights refine algorithms. Such integration is essential to overcome Rubisco’s evolutionary constraints.

• P. Crozet et al. Birth of a Photosynthetic Chassis: A MoClo Toolkit Enabling Synthetic Biology in the Microalga Chlamydomonas reinhardtii. ACS Synth Biol 7, 2074-2086 (2018).
• J. Fernandez de Cossio Diaz et al. Designing Molecular RNA Switches with Restricted Boltzmann Machines. Nature Communications 16, 11223 (2025).
• M. Huot et al. Constrained Evolutionary Funnels Shape Viral Immune Escape. 
Proc. Natl. Acad. Sci. 123, e2536956123 (2026)
• C. Malbranke et al. Computational design of novel Cas9 PAM-interacting domains using evolution-based modelling and structural quality assessment. PLoS Comput Biol 19, e1011621 (2023).
• E. Mauri, S. Cocco, R. Monasson. Mutational paths in protein-sequence landscapes: from sampling to mean-field characterization.
Physical Review Letters 130, 158402 (2023).
• P. Rochette et al. Diversity-generating retroelements for programmable targeted hypermutagenesis. Nature Biotechnology (2026) ; 10.1038/s41587-026-03078-4
• J. Thornton. Resurrecting ancient genes: experimental analysis of extinct molecules. Nat. Rev. Genet. 5, 366–375 (2004).
• W. P. Russ et al. An evolution-based model for designing chorismate mutase enzymes. Science 369, 440-445 (2020).                                                                                                                                   

Salary

The PRISM programme offers a competitive salary above the national average for PhD candidates in France to attract and support excellent researchers. Doctoral candidates will receive an approximate net monthly salary of €2,200, with additional family and mobility allowances available for eligible fellows. The salary is subject to French income tax, with the exception of the family and mobility allowances. Depending on the candidate's individual tax situation, income tax may represent approximately 2–5% of the net salary and is levied by the French tax authorities independently of the employer. To ensure consistent management and equal employment conditions across the programme, all PRISM doctoral candidates will be employed by ESPCI Paris, regardless of the host laboratory where their research is carried out.

Employer’s benefits

Remote working opportunities, access to sports and leisure activities, free access to public Paris city council’s swimming pools, access to CROUS canteen, scientific campus in central Paris, professional development programs, well-being workshops, social benefits through CNAS, partial health insurance support, and 75% support for sustainable mobility.

Name of the school of PSL

ENS - PSL (Ecole Normale Supérieure)

Research Unit

Laboratory of Physics ENS (UMR8023, LPENS)

The DC hired on RubisCO.2 will take part in the team Statistical Physics and Inference for Biology (SPIB) team of the Laboratory of Physics of ENS-PSL (LPENS), at the interface between physics, AI and biology. SPIB members are working on computational and modelling aspects of complex biological matter and systems, ranging from molecular and cell biology to neuroscience and genomics, with a strong focus on data and collaborations with experimentalists. The DC will also benefit from the vibrant environment of the LPENS, a major physics laboratory in the Paris area, with about 100 permanent researchers, 130 PhD students and 80 post-doctorates. LPENS will offer the computing power and facilities (library, office space, ...) necessary for carrying out the PhD project.

Supervision

Supervisor: Simona Cocco simona.cocco@ens.psl.eu

Co-supervisor: Rémi Monasson remi.monasson@ens.psl.eu

All projects are open PhD projects, meaning that the research plan will be further developed collaboratively by the selected doctoral candidate and the supervisors.

We seek a highly motivated DC with a strong background in physics at the Master level, including advanced statistical physics and theoretical soft matter, to contribute to the AI-driven design and analysis of RubisCO evolution. The candidate is also expected to have acquired solid training in artificial intelligence (unsupervised learning and generative models) and in data-driven modelling e.g. through internships and/or dedicated courses at the Master level. Experience with programming and high-performance computing is essential. In addition, familiarity with and/or interest for computational biology—such as sequence analysis, protein structure prediction, or evolutionary modelling—will be a significant asset, enabling the integration of experimental datasets into AI models. The ideal candidate will combine rigorous analytical skills with creativity to bridge physico-chemical-based modelling, AI, and biological data. Strong collaborative and communication skills are required to work effectively within this interdisciplinary project.

We expect DC to master advanced statistical physics techniques and concepts, including: phase transitions (order parameter, entropy-energy competition, critical phenomena), mean-field theory, dynamical processes (diffusion, master equation, path integrals, Monte Carlo Markov Chain sampling methods). Knowledge in disordered statistical physics, including spin glasses would be appreciated but is not mandatory. Solid theoretical and practical skills in machine learning are expected, covering in particular supervised learning (support vector machines and kernel methods), neural networks (stochastic gradient descent, deep architectures, physics-informed neural nets), unsupervised and generative learning (energy-based models, variational autoencoders). Experience in ML/AI programming in Python with PyTorch or equivalent will be required. We also expect DC to have acquired the basis of Bayesian inference (notions of likelihood, prior, posterior, ML and MAP estimators).