Learning developmental dynamics from lineage context
Abstract
A central goal of biology is to predict the dynamics of multicellular systems. Lineage context (the states of a cell's sisters, cousins, etc.) could inform these predictions, but how best to leverage it remains unclear. Here we introduce lineage inpainting, a generative modeling approach that uses lineage context with tree-structured transformers and stochastic interpolants to model cell state dynamics across lineage trees. Applied to a lineage-traced transcriptional atlas of mouse embryogenesis, it recovers dynamics not resolved by gene expression alone, including multiple transcriptional convergence events where distinct lineages are drawn to a common state. Despite being fit on unperturbed development, lineage inpainting can generate hypotheses about perturbation responses by simulating counterfactuals. These simulations agree in many cases with observed developmental phenotypes, including a single-cell measurement of the Tbx5 knockout response. Our results show that lineage information provides a rich source of context for learning multicellular dynamics.
Related
- Unbiased lineage tracing across the adipose organ decouples developmental origin from anatomical and transcriptomic identities
- Behavioral ontogeny in a pelagic tunicate reveals developmental maturation and context-dependent modulation
- Structurally reorganized Wmk lineage with distinct genomic context and restricted distribution in Wolbachia
- Developmental Dynamics of Bioactive Composition and Biological Activities of Methanolic Extracts from Orange (Citrus sinensis) Peel: Comprehensive Metabolite Profiling, Antioxidant Capacity, and Antibacterial Activity
- Single-nucleus transcriptomics of the Wuzhishan miniature pig kidney reveals developmental dynamics and implicates SATB2 in proximal tubule differentiation
BibTeX
@misc{gowri2026learning,
title = {{Learning developmental dynamics from lineage context}},
author = {Gokul Gowri and Nic Fishman and William N. Colgan and Minming Wang and Yusuke Kijima and Sachin A. Kumar and Aleksandr Zimin and Luke W. Koblan and Marinka Zitnik and Zachary D. Smith and Stephen Bates and Philippe Rigollet and Jonathan S. Weissman},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.40447},
eprint = {2610.40447},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.40447},
}
RIS
TY - UNPB
TI - Learning developmental dynamics from lineage context
AU - Gokul Gowri
AU - Nic Fishman
AU - William N. Colgan
AU - Minming Wang
AU - Yusuke Kijima
AU - Sachin A. Kumar
AU - Aleksandr Zimin
AU - Luke W. Koblan
AU - Marinka Zitnik
AU - Zachary D. Smith
AU - Stephen Bates
AU - Philippe Rigollet
AU - Jonathan S. Weissman
PY - 2026
DA - 2026/10/29
PB - arXiv
DO - 10.48550/arxiv.2610.40447
UR - https://arxiv.org/abs/2610.40447
AB - A central goal of biology is to predict the dynamics of multicellular systems. Lineage context (the states of a cell's sisters, cousins, etc.) could inform these predictions, but how best to leverage it remains unclear. Here we introduce lineage inpainting, a generative modeling approach that uses lineage context with tree-structured transformers and stochastic interpolants to model cell state dynamics across lineage trees. Applied to a lineage-traced transcriptional atlas of mouse embryogenesis, it recovers dynamics not resolved by gene expression alone, including multiple transcriptional convergence events where distinct lineages are drawn to a common state. Despite being fit on unperturbed development, lineage inpainting can generate hypotheses about perturbation responses by simulating counterfactuals. These simulations agree in many cases with observed developmental phenotypes, including a single-cell measurement of the Tbx5 knockout response. Our results show that lineage information provides a rich source of context for learning multicellular dynamics.
SN - 2331-8422
ER -