Learning developmental dynamics from lineage context
What it is
This paper introduces lineage inpainting, a generative modeling approach that uses tree-structured transformers and stochastic interpolants to model cell state dynamics across lineage trees. It applies this method to a lineage-traced transcriptional atlas of mouse embryogenesis to recover dynamics not resolved by gene expression alone and to simulate counterfactual perturbation responses.
The problem it solves
Standard gene expression models fail to capture the full dynamics of multicellular systems because they ignore the contextual information provided by a cell's lineage (sisters, cousins, etc.).
What you could use it for
- Predicting cell state dynamics in multicellular systems using lineage context
- Identifying transcriptional convergence events in developmental atlases
- Generating hypotheses about perturbation responses via counterfactual simulations
Why it matters
Provides a new framework for leveraging lineage information to predict the dynamics of multicellular systems, potentially improving the accuracy of developmental biology models and perturbation predictions.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.