A functional genomics atlas enhanced by convolutional neural networks facilitates clinical interpretation of disease relevant variants in non-coding regulatory elements [ChIP-seq]

We provide a novel functional genomics atlas of non-coding regulatory elements derived from human neural stem cells (NSCs), using the massively parallel reporter assay ChIP-STARR-seq. ChIP-STARR-seq was performed in NSCs using chromatin immunoprecipitation for H3K27ac, H3K4me1, YY1 and SOX2, and the generated STARR-seq plasmid reporter libraries were profiled for their enhancer activity in H9 derived NSCs and human embryonic stem cells. Subsequently a multi-omics data integration was performed, including the development of a convolutional neural network called BRAIN-MAGNET that allows to predict the effect of single nucleotide variants in non-coding regulatory elements on the enhancer activity. The various data sets generated in this study include ChIP-seq experiments in NSCs, DNA sequencing of the generated plasmid libraries, RNA-seq of transfected cells from the ChIP-STARR-seq experiments, generated ATAC-seq data sets used for data integration and the contribution scores from BRAIN-MAGNET for each nucleotide assessed in this ChIP-STARR-seq study.