Revealing RNA editing sites by using probabilistic graphical models

EDERA, AA; SANCHEZ PUERTA, MV

Buenos Aires

Congreso; XXXV Annual Meeting of the Willi Hennig Society; 2016

Willi Hennig Society

Post-transcriptional events explain the large and complex phenotypic diversity arisen from a limited number of genes. RNA editing is a post-transcriptional process by which transcripts are modified respect to the information encoded in the DNA. In flowering plants, RNA editing involves the substitution of cytidines to uridines at very specific positions of mitochondrial and plastid transcripts. RNA editing tends to increase amino acid conservation across species. Comparisons among different plant lineages reveal that the pattern of RNA editing is variable (sometimes highly homoplasious) and may lead to unexpected relationships when editing sites are included in phylogenetic analyses. Computational methods can be used to predict editing sites, but, because amino acid conservation is used as a prior knowledge, they are strongly biased to predict exclusively nonsynonymous editing sites in protein-coding genes. Previous studies have shown that synonymous editing sites are frequent and that editing also occurs in non-coding regions. We have developed a method for predicting all type of editing sites based on probabilistic graphical models. From a dataset exclusively composed of windows of nucleotides surrounding edited and non-edited cytosines from several species, statistical interactions between the target site and the other positions are exploited to automatically learn a graphical model. This model is then used to predict both synonymous and nonsynonymous editing sites at any region in a genome. Based on the predicted RNA editing sites in plant sequences, the evolution of editing sites can be inferred by standard techniques of ancestral character reconstruction to analyze editing tendencies over time.