Profile
Current Project: Neural crest (NC) cells are a vertebrate-specific multipotent cell population whose evolutionary origins remain incompletely understood. Although non-vertebrate chordates possess cell populations with partial neural crest–like properties, the cis-regulatory mechanisms underlying the emergence of neural crest identity remain unclear. In this project, we combine deep learning–based chromatin accessibility models within silico evolution to investigate how neural crest enhancers may evolve across chordates. Using convolutional neural network (CNN) models trained on chicken cranial NC ATAC-seq data, we evolved synthetic enhancers from random DNA sequences. In vivo assays in chick embryos demonstrated that active neural crest enhancers can emerge after 15 rounds of evolution. Large-scale evolution of random genomic sequences showed that in silico evolution is strongly influenced by the starting sequence. We are currently using this property to identify evolvable regions in the genomes of zebrafish and non-vertebrate chordates and investigate the motif features in the evolvable regions.