Regulon Velocity-Based Single-Cell Dynamics for Interpretable Modelling of CD8 T-Cell Fate Commitment

How CD8 T cells commit to distinct differentiation states is a central question in immunology and an important challenge for cell engineering. In this seminar, I will present a framework under development that projects gene expression and RNA velocity into an interpretable regulon activity space, enabling continuous dynamical modeling over transcription factor regulatory programs rather than gene-expression coordinates alone. Using chronic LCMV-specific CD8 T cells as a model system, I will show how this approach supports the reconstruction of continuous vector fields, the identification of fixed points, and the characterization of saddle-local neighborhoods that are especially informative for fate commitment. I will focus in particular on a saddle-like neighborhood along the Tpex–Texint differentiation axis, where local dynamical analysis reveals branch-defining unstable modes and opposing transcription factor modules associated with commitment bias. More broadly, this work illustrates how differentiable regulon-space dynamics can connect single-cell state transitions to mechanistic regulatory interpretation, with potential applications in regulator prioritization, reprogramming prediction, and CD8 T-cell engineering for cancer immunotherapy.