Related Experiment Video
Updated: May 21, 2026
07:59
An Optogenetic Method to Control and Analyze Gene Expression Patterns in Cell-to-cell Interactions
Published on: March 22, 2018
Inferring stochastic dynamics by biophysical Neural ODE using single-cell transcriptomics
Jingyu Dou1,2, Wentao Lyu2, Feng Chen1
1Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China.
Nature Communications
|May 19, 2026
Summary
DynNet, a novel deep learning method, infers gene regulatory dynamics for cell fate decisions from single-cell data. It overcomes limitations of existing models, accurately reconstructing cell state transitions and developmental trajectories.
Area of Science:
- Computational Biology
- Systems Biology
- Genomics
Background:
- Single-cell RNA sequencing (scRNA-seq) enables cellular heterogeneity analysis but inferring dynamic processes from static data is challenging.
- Existing computational models for cell fate decisions involve a trade-off between biological realism and interpretability.
- Understanding gene regulatory dynamics is crucial for dissecting cell fate decisions and developmental processes.
Purpose of the Study:
- To develop a deep learning framework, DynNet, that integrates Neural Ordinary Differential Equations (Neural ODEs) with biophysical models and prior knowledge.
- To accurately infer stochastic gene regulatory dynamics governing cell fate decisions from scRNA-seq data.
- To overcome the limitations of current mechanistic and data-driven approaches in modeling cellular dynamics.
Main Methods:
- DynNet integrates Neural ODEs with biophysical models and prior knowledge of gene expression.
- The method learns the stochastic dynamics of gene regulatory systems for cell fate decisions.
- Benchmarking was performed on synthetic datasets and real biological data (hepatocyte differentiation and EMT).
Main Results:
- DynNet successfully infers stable cell states and reconstructs dynamical trajectories from static scRNA-seq data.
- The method characterizes multi-stable cell fate transitions and quantifies transition probabilities.
- Application to hepatocyte differentiation and EMT data revealed developmental trajectories, cell fate landscapes, and critical gene regulations.
Conclusions:
- DynNet provides a powerful, interpretable deep learning approach for inferring gene regulatory dynamics and cell fate landscapes from scRNA-seq data.
- The method advances our understanding of cellular heterogeneity and cell fate decisions in developmental and disease contexts.
- DynNet offers a promising tool for dissecting complex biological processes at the single-cell level.
