Dynamic Interrogation of Stochastic Transcriptome Trajectories (DIST2)

Abstract

New methods in genomics allow the tracking of single cell transcriptome across tens of thousands of genes for hundreds of cells dynamically changing over time. These advancements open new computational problems and provide opportunity to explore new solutions to the interrogation of the transcriptome data in humans and in animal models. Common data analysis pipelines include a dimensionality reduction step to facilitate visualizing the data in two or three dimensions, (e.g. using t-distributed stochastic neighbor embedding (t-SNE)). Such methods reveal structure in high-dimensional data, while aiming at accurately representing global structure of the data. A potential pitfall of some methods is gross data loss when constraining the analyses to gene space data that is not asynchronously changing from day to day, or that express more stable variability of some genes relative to other genes.