Bagherian, M.; Harris, G.; Sathishkumar, P.; Lloyd, J. P. B.

Proper identification of start codons is critical for accurate annotation of open reading frames (ORFs) and prediction of nonsense-mediated mRNA decay (NMD) targets. We demonstrate that reference transcriptomes annotations frequently select the longest ORF rather than the authentic start codon, leading to incorrect assessment of NMD features. Our revised ORF annotations significantly improve NMD target identification, and prevents computational hallucinations in protein databases.