ATAC-seq of frozen H1 Human Embryonic Stem Cells (hESC)

We performed Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) to profile genome-wide chromatin accessibility in the human H1 embryonic stem cell (ESC) line. We used this data to train a deep learning model called ChromBPNet which can accurately predict base-resolution accessibility profiles as a function of DNA sequence, while accounting for and correcting biases due the sequence preferences of the Tn5 transposase used in ATAC-seq. We interpreted the models to identify globally predictive transcription factor (TF) motifs, individual predictive motif instances in all accessible regions and Tn5-bias corrected canonical footprints of TFs at these predictive motifs. Overall design: Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) was used to profile genome-wide chromatin accessibility in the human H1 embryonic stem cell (ESC) line. We then used this data to develop ChromBPNet, a deep learning model capable of predicting DNA sequence-based base-resolution accessibility profiles, correcting biases from Tn5 transposase in ATAC-seq, and revealing globally predictive transcription factor motifs and their specific instances, along with Tn5-bias-corrected canonical footprints for TFs at these motifs.