Doctoral Candidate Virginia Tech Christiansburg, Virginia, United States
Disclosure(s):
Tian Xu: No financial relationships to disclose
Introduction/Rationale: Modeling the temporal evolution of the immune system remains a central yet unsolved challenge in computational immunology. While single-cell transcriptomics provides unprecedented cellular resolution, most analyses are confined to static distributions that lack true temporal context. We introduce Single-Cell Brownian Bridge (SCBB), a novel stochastic–deep learning framework that infers the hidden temporal dynamics of immune cell evolution from a single timepoint.
Methods: SCBB establishes a mathematical bridge between static molecular snapshots and continuous biological time. By representing immune state transitions as Brownian bridge diffusions constrained by biologically meaningful endpoints, and embedding these within a forward Markov process, SCBB enables reverse-time learning—the ability to reconstruct developmental or pathological immune trajectories backward from their mature states. This fusion of stochastic process theory and generative modeling transforms static scRNA-seq data into dynamic maps of immune evolution.
Results: Applied to human and murine immune datasets, SCBB reveals latent differentiation hierarchies, transition probabilities, and emergent attractor states underlying tolerance, aging, and autoimmunity.
Conclusion: By mathematically grounding biological time within probabilistic manifolds, SCBB represents a conceptual shift from snapshot-based inference to stochastic temporal reconstruction, opening a new avenue for decoding immune system evolution across lifespan and disease.
