Virtual Embryo

Virtual Embryo is an open, AI-native platform working towards a predictive virtual embryo — a computational model that integrates single-cell and spatial genomics with AI to simulate mammalian embryogenesis across scales, from molecular trajectories to whole-organism morphogenesis. The current site is the accompanying data and knowledge layer: a unified atlas plus a knowledge graph that both humans and LLM agents can query. Mouse is the current focus; scope extends to other species — including human — as the data layer matures, with congenital disease mechanisms as a translational target.

The atlas brings together resources from many groups. We gratefully acknowledge the Edinburgh Mouse Atlas (Baldock group, University of Edinburgh) for the Kaufman histology plates, 3D embryonic reference models, and the EMAPA anatomy ontology that anchor the mouse layer. Beyond that foundation, Virtual Embryo integrates contemporary single-cell atlases, spatial-transcriptomics datasets, drug / phenotype / disease ontologies, and an LLM-driven paper-extraction pipeline that adds new claims to the knowledge graph.

The emerging human layer references established human-development resources rather than hosting their data. We gratefully acknowledge the HDBR Atlas (Human Developmental Biology Resource, Newcastle University & UCL, funded by the MRC and Wellcome): its 3-D model thumbnails are shown on the Carnegie-stage timeline under CC BY-NC-SA 4.0 with attribution. We also reference the 3D Atlas of Human Embryology (de Bakker et al., Science 2016) and the Virtual Human Embryo with links out to the originators.

Initiated by the Qiu Lab at Stanford University. Supported by a Moonshot grant from the Laude Institute.