Deciphering the evolution of gene regulatory networks and enhancer architectures in neuronal cell types using single-cell deep learning and comparative genomics

Genomic enhancers are key players in gene regulatory networks (GRNs), underlying the differentiation of cell types. On the one hand, genomic regulatory blocks (GRBs) reflect conservation of enhancer-gene pairs for key developmental genes. On the other hand, rewiring of GRNs through the birth, erosion, and repurposing of enhancers is a key driver for evolution across mammals, including the emergence of the neocortex. Recent single cell ATACseq (scATACseq) methods provide the possibility to examine enhancer usage in distinct cell types, across species. Deep learning approaches have successfully been applied for the analysis of scATACseq data, yielding significant insight into enhancer architectures. However, there is little knowledge about how evolutionary changes in enhancers lead to rewiring of GRNs, and how that impacts different neuronal cell types. I will utilize scATACseq data and combine deep learning with comparative genomics methods; and develop a Graph Neural Network method to jointly learn GRNs and enhancer architectures, characteristic for neuronal cell types, and investigate how these GRNs were rewired during evolution through rearrangements of GRBs, gain, erosion, and repurposing of enhancers using comparative genomics. This project will provide insight into the evolutionary time at which different neuronal cell types evolved and how rewiring of enhancers relates to the evolution of neural cell types in vertebrate and invertebrate.