Correction of gene model annotations improves isoform abundance estimates: the example of ketohexokinase (Khk)

• DOI: 10.12688/f1000research.17082.1
• Publication Date: 2018-12-19T13:20:07Z
• Authors: Christophe D. Cha...; Tanja Eberhart; Ilaria Guccini; Wilhelm Krek; Werner J. Kovacs
• Keywords: 0301 basic medicine; Models, Genetic; 0206 medical engineering; Datasets as Topic; Molecular Sequence Annotation; 02 engineering and technology; Introns; Fructokinases; Mice, Inbred C57BL; Alternative Splicing; 03 medical and health sciences; Organ Specificity; Animals; Protein Isoforms; Research Article

Next generation sequencing protocols such as RNA-seq have made the genome wide characterization of transcriptome a crucial part many research projects in biology. Analyses resulting data provide key information on gene expression and certain cases exon or isoform usage. The emergence transcript quantification software Salmon has enabled researchers to efficiently estimate expressions across while tremendously reducing necessary computational power. Although overall estimations were shown be accurate, quantifications appear more challenging task. Low levels uneven insufficient coverage reported potential explanations for inconsistent estimates. Here, through example ketohexokinase ( Khk) mouse, we demonstrate that use an incorrect annotation can also result erroneous results. Manual correction input Khk model provided much accurate estimation relative when compared quantitative PCR (qPCR measurements). In particular, removal unexpressed retained intron proper adjustment 5' 3' untranslated regions both had strong impact Finally, observed better concordance between datasets strategies relying newly generated annotations. These results highlight importance models annotations correct reassert need orthogonal methods confirm important findings.