Skip to main content
Go to the home page of the European Commission (opens in new window)
English en
CORDIS - EU research results
CORDIS

Reverse-engineering the development of embryos with physics-informed machine learning

Project description

Reverse engineering a virtual embryo sheds light on embryogenesis

The formation and development of an embryo from a fertilised egg – a single cell – is a formidable task directed by nature with exquisite precision and control. Aside from cell division and differentiation, the cells must find their way to the appropriate places to form the tissues and organs that will sustain the new organism. Chemical signalling historically held the spotlight as the key driver of all this change, but over the last several decades, it has become clear that mechanical forces generated by the living cells are equally relevant. The EU-funded DeepEmbryo project will apply advanced in silico methods to create a virtual embryo from experimental data and to test predictions about the interplay of mechanical forces and signalling during embryogenesis.

Objective

Embryogenesis is archetypal of a self-organized process, where the emergence of a complex structure stems from the interaction of its elementary parts. Progress in imaging and molecular genetics allow us to delve into embryos at unprecedented spatiotemporal resolutions, but extracting biophysical information from this complex multidimensional data is a highly technical challenge. As a result the principles of multicellular self-organization remain far from understood. DeepEmbryo proposes to fill this gap by pioneering the use of deep learning to reverse-engineer early embryo development directly from high-resolution 3D microscopy movies. Focusing on four animal groups (mammals, ascidians, nematodes and annelids), the project will combine physical modeling and machine learning to tackle three fundamental questions from a unique transversal perspective: Q1 What are the forces shaping early embryos? Using convolutional neural networks, I will develop an automated method to directly infer cell forces from membrane-labeled images of embryos. Q2 How do cells coordinate forces, division and signaling? Regarding cells as dynamical systems, I will model them with minimal neural networks and design a multi-agent embryo model able to learn by reinforcement the fundamental feedback controls between mechanics and fate. Q3 What principles ensure developmental robustness? Using deep generative models, I will infer intra-specie developmental variability to identify robust developmental traits and mechanisms. Using dropout techniques as virtual analog to genetic knockout, I will produce experimentally testable new predictions, refining my inaugural virtual embryos. Pioneering a new field at the frontier of developmental biology, artificial intelligence and physics, DeepEmbryo will uncover the fundamental engineering principles of early embryogenesis, with far-reaching implications in multi-agent modeling, evolutionary biology, physical inference and tissue engineering.

Fields of science (EuroSciVoc)

Keywords

Programme(s)

Topic(s)

Funding Scheme

ERC-STG - Starting Grant

See all projects funded under this funding scheme

Call for proposal

(opens in new window) ERC-2020-STG

See all projects funded under this call

Host institution

CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE CNRS
Net EU contribution
€ 1 957 751,00
Address
RUE MICHEL ANGE 3
75794 Paris
France
Region
Ile-de-France Ile-de-France Paris
Activity type
Research Organisations
Links
Total cost
€ 1 957 751,00

Beneficiaries (1)