|LEON-NOVELO, LUIS - University Of Texas Health Science Center
|GRAZE, RITA - Auburn University
|MCINTYRE, LAUREN - University Of Florida
|MARRONI, FABIO - University Of Udine
Submitted to: G3, Genes/Genomes/Genetics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/15/2017
Publication Date: 3/1/2018
Publication URL: https://handle.nal.usda.gov/10113/6472315
Citation: Leon-Novelo, L., Gerken, A.R., Graze, R.M., McIntyre, L.M., Marroni, F. 2018. Direct testing for allele-specific expression differences between conditions. G3, Genes/Genomes/Genetics. 8:447-460. https://doi.org/10.1534/g3.117.300139.
Interpretive Summary: Differences in genes among individual organisms contribute to variation in observable traits, but differences in the process of translating information from genes to finished products (gene expression) can also cause variation in gene expression as observable traits. Allelic imbalance occurs when changes to gene expression are caused by nearby gene regions. Tracking these regulatory changes can provide insight into disease states and environmentally induced traits, but evaluating allelic imbalance is difficult. Using a novel model for analyzing differences in allelic imbalance, we have shown allelic imbalance in 20% of the gene regions using a model system consisting of a panel of 68 strains of Drosophila melanogaster. Our model also corrects for bias using variability in the DNA sequencing technology, whereas previous models used filtering to throw out oddities. Using simulations, we show that the new model has a low error rate under several scenarios, and is robust to large differences in sequencing coverage. Our new model can also test the differences in regulation between two different environments and we find that only 7% of regulatory regions show differences in mated and virgin flies. The regulatory regions that differed between mated and virgin flies contain genes involved in immune response, while regions that were stable included genes which would be key to maintaining gene expression in regions that if affected would cause the overall function of the organism to fail. These results show that regulatory variation is common in natural populations, but that genetic regulation is robust. This conservative model of allelic imbalance estimation can be used to explicitly estimate differences in genetic regulation in disease and healthy states, possibly pinpointing areas of the genome that are influential in driving diseases and natural variation.
Technical Abstract: Genetic differences in cis regulatory regions contribute to the phenotypic variation observed in natural and human populations, including beneficial, potentially adaptive, traits as well as disease states. The two alleles in a diploid cell can differ in their allele-specific expression leading to allelic imbalance (AI). AI indicates functional variation in cis regulatory regions. Detecting cis differences using AI based approaches has become widespread, yet there is no formal statistical methodology that tests whether AI differs between conditions or environments. Here we present a novel model for analyzing differences in AI across treatments based on Bayesian credible intervals. We do not rely on filtering to account for bias, but instead incorporate potential bias into the modeling process. The new model also explicitly accounts for depth of coverage, a critical factor as coverage affects power for detection of AI and can differ between conditions. Using simulations, we show that the new model has low type I and II error under several scenarios, and is robust to large differences in sequencing coverage between conditions. We reanalyze RNA-seq data from a Drosophila melanogaster population panel, with F1 genotypes, to compare levels of AI between mated and virgin female flies. We find that AI is generally robust to environmental variation but that some differences are observed. We tested AI in 13,898 exons belonging to 6,840 genes and 68 F1 genotypes, for a total of 169,842 tests. We identified AI in 20% of tests, and environmental variation of AI in 7% of tests. From these data we derive novel insight into both the genes showing regulatory variation in response to a given condition and those which are unusually stable in their expression patterns across conditions.