A Flow Artist for High-Dimensional Cellular Data
This addresses visualization challenges in dynamic biological data like single-cell transcriptomics, though it appears incremental as it builds on existing embedding techniques by combining coordinate and velocity information.
The authors tackled the problem of embedding high-dimensional point cloud data with associated velocity information, presenting FlowArtist which jointly embeds points while learning a vector field to better separate and visualize velocity-informed structures. They demonstrated results on toy datasets and single-cell RNA velocity data.
We consider the problem of embedding point cloud data sampled from an underlying manifold with an associated flow or velocity. Such data arises in many contexts where static snapshots of dynamic entities are measured, including in high-throughput biology such as single-cell transcriptomics. Existing embedding techniques either do not utilize velocity information or embed the coordinates and velocities independently, i.e., they either impose velocities on top of an existing point embedding or embed points within a prescribed vector field. Here we present FlowArtist, a neural network that embeds points while jointly learning a vector field around the points. The combination allows FlowArtist to better separate and visualize velocity-informed structures. Our results, on toy datasets and single-cell RNA velocity data, illustrate the value of utilizing coordinate and velocity information in tandem for embedding and visualizing high-dimensional data.