Time-variant clustering model for understanding cell fate decisions

Significance Clustering, in essence, was a tool for describing spatial characteristics in the objects of concern. However, from social to biological studies, researchers’ interests in temporal characteristics often rival their interests in spatial features. Previous efforts to incorporate time information into clustering primarily relied on coding the time information into similarity metrics, thus reducing the problem into the classical paradigm of spatial clustering. The limitation is that the clustering outcomes are often time invariant. Here, we initiate a class of statistical methods that simultaneously infer spatial and temporal groupings. Such methods explicitly model the time dependencies of clustering indices over time. Our method inferred three genes to be associated with the earliest cell fate decision, which was corroborated by experimental validations.