Deep-learning based analysis of microtubule network in cells

Host laboratory:

The internship will be carried out within the Rennes Institute of Genetics and Development (IGDR, UMR 6290), and more specifically in the CeDRE team "Reverse Engineering of the Cell Division". Our team – which has the specificity of being an interdisciplinary team made up of specialists in biology, physics, image analysis, artificial intelligence and bioinformatics – studies cell division robustness using a cell biophysical approach. For this, we use the first division of the Caenorhabditis elegans nematode, an established model of cell division, and then validate our findings in Human cells. We aim to understand the robustness of cell division by studying and modelling the biophysical and mechanical interactions between the molecular actors of mitosis, which are the microtubules and their regulators, as well as the molecular motors.

Internship project:

Astral microtubules are dynamic and semi-flexible fibres. Their dynamics were studied in vivo, which revealed their major role in cell division, and in particular concerning astral microtubules in the correct positioning of the mitotic spindle [1-2]. In contrast, the mechanical characteristics of microtubules in vivo are not well known, and their contribution to cell division remains to be identified. Microtubule curvatures can serve as proxy for revealing perturbations in microtubule rigidity. The team as already developed a Deep-learning-based tool capable of segmenting curvilinear structures, including microtubules, in noisy 2D images. For that we build a U-Net architecture that combines residual networks and attention mechanisms [3]. In the present internship, we aim add developing a similar tool that will succeed to segment microtubules in 3D images, and even to go one step further by directly extracting microtubule curvatures. By developing deep-learning based tools for 3D microscopy data, the student will help understand how changes in microtubule rigidity influence cell division, while gaining hands-on experience in designing and applying state-of-theart deep-learning methods for complex biomedical image analysis.

References:
[1] Bouvrais, H., et al., Biophysical Journal, 2018. 115(11): p. 2189- 2205.
[2] Bouvrais, H., et al., EMBO reports, 2021. 22(5): p. e50770.
[3] Ait Laydi, A., et al., arXiv preprint arXiv:2507.07800, 2025

Duration: From 2 to 6 months

Period: Between January and July 2026.

Contact:

Dr Sidi Mohamed Sid’El Moctar (post-doctoral fellow) 

Dr Hélène Bouvrais (CNRS Researcher) 

Web site: https://igdr.univ-rennes.fr/en/cedre-lab

Skills:

- Required: Solid programming skills in Python; Basic understanding of deep learning architectures (e.g., CNNs, U-Net).

- Preferred: Interest in biological applications and ability to work in an interdisciplinary environment; Knowledge in image analysis.

Skills that will be acquired: Image analysis, deep-learning, microscopy, Improvement in English, teamwork skills.