Mots-Clés

Computational biology, bioinformatics, machine and deep learning, generative models, NGS data analysis, single-cell genomics, gene regulation, stem cells and cell differentiation.

Description

Biophysics-informed interpretable deep learning to analyze multimodal single-cell sequencing data

Postdoc position – Molina & Kim & Gradwohl Labs – IGBMC

Project summary: Biology is going through an incredible revolution: single-cell genomics allows to characterize all cells types in the human body gaining a systematic understanding of collective cell function in health and disease. This poses exciting challenges in computational biology as new approaches are required to extract valuable information and produce reliable predictions. Deep learning has been proven to be a very powerful tool to accomplish these tasks. However, the black-box nature of deep learning methods hinders the interpretability of the models and limits their predictability power. To overcome this problem, we develop interpretable models combining generative deep learning with biophysical principles by integrating prior biological knowledge of molecular interactions relevant in shaping gene expression dynamics. We train our models in large-scale single-cell multiomic datasets and identify key regulators driving specific biological processes such as cell-cycle progression and cell-fate determination1,2. Moreover, we are able to infer gene-specific non-linear response functions that are able to capture complex combinatorial regulations. Finally, we used optimal transport theory to predict the optimum regulatory path to promote a transition from one particular cell type to another. We offer a postdoc position for a computational biologist to join the team and get involved in this project and develop interpretable generative models to analyze multimodal single-cell genomic data. The candidate will work in close collaboration with the Gradwohl and Kim teams analyzing newly generated data to study cell differentiation and nuclear fate establishment in normal and pathological conditions using gut and muscle as organ models2,3. We can offer up to 4 years of salary upon positive interim evaluations, but the ideal candidate will be encouraged and supported to apply for postdoctoral fellowships with the possibility to eventually apply to a permanent position at CNRS, INSERM or University of Strasbourg.

Required skills: The ideal candidate should have a PhD degree in Computational Biology, Bioinformatics, Machine Learning, Data Science or similar. Proven experience applying and developing deep learning tools for the analysis of NGS data is expected. Previous experience analyzing single-cell sequencing experiments will be a plus. A strong background in programming is expected. A high motivation and a good capacity to work in a multidisciplinary team are also important. English is the working language of the team and the IGBMC.

Key words: Computational biology, bioinformatics, machine and deep learning, generative models, NGS data analysis, single-cell genomics, gene regulation, stem cells and cell differentiation.

References

[1] https://www.nature.com/articles/s41467-022-30545-8

[2] https://www.biorxiv.org/content/10.1101/2022.04.01.486696v1

[3] https://www.nature.com/articles/s41467-020-20064-9

Contact

Nacho Molina (nacho.molina@igbmc.fr)  https://www.igbmc.fr/en/recherche/teams/stochastic-systems-biology-of-gene-regulation

Gérard Gradwohl (gradwohl@igbmc.fr) https://www.igbmc.fr/en/igbmc-1/departments/developmental-biology-and-stem-cells

Minchul Kim (kimm@igbmc.fr) https://www.igbmc.fr/equipes/biologie-des-cellules-syncytiales