Levy-Driven Single Cell Omics Regulatory Dynamics (Levy-scORD) Model
Please login to view abstract download link
We formulate Levy-scORD, a controlled jump-diffusion model of gene-regulatory dynamics coupling continuous transcription-factor and mRNA expression profiles with discrete cell-state transitions. High-dimensional scRNA-seq is treated through an autoencoder latent embedding on which a forward--backward sweep method realises stochastic maximum-principle optimality conditions. On Neftel and GBmap glioblastoma datasets, classification of the terminal latent state recovers cell-type structure with high fidelity, indicating that L\'evy-driven latent control is a mathematically consistent and computationally scalable route to modelling phenotypic plasticity.
