DeTECTOR: GENE TRANSLATION IN EUKARYOTES, A DIAGNOSTIC TOOL

Abstract

Gene prediction pipelines are often affected by a high false positive ratio. This is true in particular for splicing variant detection. We describe here a diagnostic tool (DeTECTOR: Detection Tool of Elements Controlling Translation of Open Reading frames), based on a neural network and aimed at classifying transcripts into three major classes: coding, pseudogenes and non-translated RNA. It evaluates the presence of transcriptional regulative elements, together with the properties of the putative translated products. It reaches good accuracy in detecting non-translated RNAs, while its performance in discriminating pseudogenes from coding transcripts is less satisfactory.