Reconstruction of Cellular Trajectories with Biologically Interpretable States

A Latent Cell-State Model reconstructs cellular evolution from scRNA-seq by inferring biologically meaningful hidden cell states and generating high-quality gene expression at unobserved timepoints.

• Why it matters: scRNA-seq yields only destructive, discrete snapshots, and existing trajectory models do not recover each cell’s hidden state needed to estimate expression between measured times.
• What they did: The approach treats cell state changes as a Markov chain and models gene expression at unobserved timepoints as dependent on state, time, and a latent variable capturing high-dimensional expression.
• The result: The inferred states are biologically meaningful and the model produces high-quality expression profiles at unobserved timepoints, enabling more accurate continuous trajectory reconstruction.

AI summary of the abstract · may contain errors · not medical advice · read the paper.