Original title: scNODE: Generative Model for Temporal Single Cell Transcriptomic Data Prediction
Authors: Jiaqi Zhang,Erica Larschan,Jeremy Bigness,Ritambhara Singh
In this article, the authors discuss the challenges associated with studying gene expression programs in individual cells over time using single-cell RNA sequencing (scRNA-seq) technology. The current approach of sampling gene expression at discrete time points leads to information loss between these points and hinders downstream developmental analysis. To overcome this problem, the authors propose a new model called scNODE.
scNODE is an end-to-end model that can simulate and predict realistic single-cell samples at any time point. It combines a variational autoencoder with neural ordinary differential equations to predict gene expression in a continuous and non-linear latent space. The authors also add a regularization term to capture the overall dynamics of cell development in the latent space, making the learned representation informative and interpretable.
The authors evaluate scNODE using six real-world scRNA-seq datasets and find that it outperforms existing methods in terms of predictive performance. They also show that scNODE can learn an interpretable latent space and accurately recapitulate cell-type developmental trajectories.
No competing interests were declared by the authors.
Original article: https://www.biorxiv.org/content/10.1101/2023.11.22.568346v1