Marker genes reveal dynamic features of cell evolving processes.

Motivation: Embryonic cells finally evolve into various types of mature cells, where cell fate determinations play pivotal roles, but dynamic features of this process remain elusive. Results: We analyze four single-cell RNA sequencing datasets on mouse embryo cells, mouse embryonic fibroblasts, human bone marrow, and intestine organoid. We show that key (high expression) genes of each organism exhibit different statistical features and expression patterns before and after branch, e.g. for mouse embryo cells, the mRNA distribution of gene Gata3 is bimodal before branch, unimodal at branching point and trimodal for one branch but bimodal for the other branch. Moreover, there is a distribution mode such that it is the same before and after branch, and this fact would account for maintenance of the genetic information in a complex cell evolving process. Machine learning reveal that along the cell pseudo-time trajectory, the strength that one key gene regulates another is fundamentally increasing before branch but is always monotonically increasing after branch; burst size and frequency of key genes are always monotonically decreasing before branch but monotonically increasing for one branch and monotonically decreasing for another branch. Our results unveil the essential features of dynamic cell processes and can be taken as a supplement for accurately screening marker genes of cell fate determination on basis of the existed methods. Availability and implementation: The implementation of CFD is available at https://github.com/cellwj/CFD and the preprocessed data is available at https://zenodo.org/records/14367638.Cell fate determination, single-cell RNA sequencing data, marker gene, cell process, developmental branch.