School of Informatics and Computing Menu

Seyedsasan Hashemikhabir, PhD candidate: Uncovering the RNA editing landscape in Glioblastoma to identify a compendium of prognostic editing alterations to classify cancer subtypes

Friday, April 21 at 12:30 p.m. in IT 252


RNA editing is increasingly appreciated as an important posttranscriptional regulatory mechanism in mammals. Adenosine deaminases that act on RNA (ADARs) are the enzymes that catalyze adenosine (A) to inosine (I) editing events. Human brain RNA is reported to have the highest number of editing events. Many neurotransmitter receptors and ion channels undergo editing within exonic regions, which generates a different protein that that encoded by the genome. In addition, ALU repeats in introns and untranslated regions of brain mRNAs are often targeted by editing events and result in altered splicing and post-transcriptional gene regulation. In this study, we developed a comprehensive framework for high confidence identification of RNA editing events in ~160 glioblastoma patient samples from The Cancer Genome Atlas (TCGA) and compared them with the frontal cortex samples from the Genotype-Tissue Expression Project (GTEX) as the normal control. This framework enabled the identification of thousands of very high confidence A to I edited sites, most of which are patient specific and located in non-coding genomic regions. ADAR1, the primary A-to-I editing enzyme, is expressed significantly lower in the normal samples compared to the glioblastoma, however, ADAR2 and ADAR3 expressions are elevated in the normal tissues. We also show that the ADAR1 expression levels are significantly correlated with the frequency of the edited sites in glioblastoma patients, however editing levels can not be predicted by ADAR1’s levels using a simple linear correlation. We identified ~780 differentially edited events between tumor and normal samples, of which a third of them were found to be associated with variability in the underlying gene expression levels (Cis-RNA Editing QTL) and the majority of them are enriched in 3’ UTR regions. Although, differentially edited events frequently exhibit decreased editing levels in the cancer samples, their underlying gene expression was significantly higher in cancer compared to the normal controls. Finally, we selected the top 44 edited sites contributing to the glioblastoma subtype classification using the High Dimensional Discriminant Analysis (HDDA) as a feature selection method. The editing levels of the selected sites enables us to accurately classify the glioblastoma subtypes in the patients (balanced accuracy per subtype = ~0.75).

About Sasan Hashemi

Sasan is a PhD candidate in bioinformatics at the School of Informatics and Computing, IUPUI. He received his Bachelor’s degree in Computer Science from Urmia University followed by a Master’s degree from the Middle East Technical University. He started his PhD at IUPUI in 2013. Sasan’s primary research interests are to understand the structure of post-transcriptional regulatory networks across normal tissues and how they are altered across cancer genomes. He is also interested in developing computational algorithms and pipelines that facilitate the downstream analysis of the high-throughput sequencing data.