Abstract:
Pineoblastoma is rare but aggressive pediatric brain tumor of the pineal region. Currently, there is no gold standard of care for treating pineoblastoma because key aspects that differentiate pineoblastoma from other well described brain tumors at the genetic and epigenetic levels are still lacking.
In this project, we utilized a pineoblastoma transgenic mouse model (Cyclin D1 p53-/-) to conduct a time series experiment by collecting pineal gland tissues at key time points representing the proliferating, premalignant, and tumorigenic stages of pineoblastoma. We performed several next generation sequencing experiments to interrogate the genome-wide gene expression (RNA-seq), chromatin accessibility (ATAC-seq), and active enhancers marked by the H3K27ac histone mark (ChIP-seq) and integrated our data using a bioinformatics and machine learning approach.
We identified time point specific up and down-regulated gene clusters with distinct functional properties in pineoblastoma and discovered that key structural events of potential oncogenic nature mark the onset and progression of the disease. We also described the dynamic enhancer landscape of pineoblastoma at the tumorigenic stage and identified key transcription factors involved in epigenetic regulation of gene expression. To address the issue of missing ChIP-seq data at certain time-points, we developed a deep learning predictive model, TempoMAGE, that learns from the data generated from the different experimental assays to predict the presence or absence of H3K27ac histone mark in open chromatin.
