Predicting cell-type specific combinatorial binding of neuronal transcription factor network by deep learning

The adult Drosophila peripheral nervous systems (PNS) are mechano-sensory organs built according to a strict lineage plan directed by regulatory networks, and are an excellent model system for cell specification. Decades of genetic studies revealed that cell-type specific transcription factors (TFs) collaborate to specify PNS fates. However, much remains to be understood on how high-level TFs execute their function via target networks. We generated genomic binding data for the 10 PNS TFs using ChIP-seq and ChIP-nexus assays. By combining them with single cell accessibility data (sciATAC-seq) and RNA-seq data (scRNA-seq), we are building deep learning models that predict cell-type specific chromatin accessibility from TF binding motif combination and precise gene expression patterns and identifies cell-type from the combination of genomic and epigenomic input features.

This project is a collaboration with Qi Dai.