Inferring Dynamic Cellular Trajectories and Underlying Cellular Regulatory Networks with Neural and Graph ODE Models Part 1

Smita Krishnaswamy
Yale University

This two-part tutorial. The first part will cover our data diffusion geometry-preserving PHATE method for embedding cellular data into a low dimensional state space, and then to infer cellular trajectories within that space using our biologically principled neural ODE framework called MIOflow (manifold interpolating flows). In the second part, we will cover our graph ODE framework called RITINI (Regulatory Interaction Network Inference) to infer regulatory mechanisms underlying these dynamic cellular trajectories. We will also cover a case study in breast cancer, showing how cells transform from cancer stem cell states to aggressive metastatic states, and how these methods could be used to discover potential state-gating targets.


Back to Long Programs