Identifying genome-wide transcription units from histone modifications using EPIGENE

With the successful completion of the human genome project and the rapid development of sequencing technologies, transcriptome annotation across multiple human cell types and tissues is now available. Accurate transcriptome annotation is critical for understanding the functional as well as the regulatory roles of genomic regions. Current methods for identifying genome-wide active transcription units (TUs) use RNA sequencing (RNA-seq). However, this approach requires large quantities of mRNAs making the identification of highly unstable regulatory RNAs (like microRNA precursors) difficult. As a result of this complexity in identifying inherently unstable TUs, the transcriptome landscape across all cells and tissues remains incomplete. This problem can be alleviated by chromatin-based approaches due to a well-established correlation between transcription and histone modification. Here, I present EPIGENE, a novel chromatin segmentation method for identifying genome-wide active TUs using transcription-associated histone modifications. Unlike existing chromatin segmentation approaches, EPIGENE uses a constrained, semi-supervised multivariate Hidden Markov Model (HMM) that models the observed combination of histone modifications using a product of independent Bernoulli random variables to identify the chromatin state sequence underlying an active TU. Using EPIGENE, I successfully predicted genome-wide TUs across multiple human cell lines. EPIGENE predicted TUs were enriched for RNA Polymerase II (Pol II) at the transcription start site (TSS) and in gene body indicating that they are indeed transcribed. Comprehensive validation using existing annotations revealed that 93% of EPIGENE TUs can be explained by existing gene annotations and 5% of EPIGENE TUs in HepG2 can be explained by microRNA annotations. EPIGENE predicted TUs more precisely compared to existing chromatin segmentation and RNA-seq based approaches across multiple human cell lines. Using EPIGENE, I also identified 232 novel TUs in K562 and 43 novel cell-specific TUs in K562, HepG2, and IMR90, all of which were supported by Pol II ChIP-seq and nascent RNA-seq evidence