Reference protein-coding transcripts of human genes annotated using long-read transcriptome datasets.

Area of Science:

• Genomics
• Transcriptomics
• Bioinformatics

Background:

• Next-generation sequencing (NGS) suggests many alternatively spliced transcripts, but short-read data may overestimate this due to isoform ambiguity.
• Accurate identification of tissue-specific expression profiles is essential for determining functional, translated peptide products.

Purpose of the Study:

• To identify the most highly expressed protein-coding transcripts (Ref-Tx) in human genes using long-read sequencing data.
• To assess the accuracy of dominant transcript isoform identification compared to existing annotation datasets.
• To develop a bioinformatics tool for visualizing tissue-specific expression of dominant transcripts.

Main Methods:

• Utilized the GSE192955 long-read nanopore sequencing dataset from 30 normal human tissues.
• Identified 18,094 dominant representative protein-coding transcripts (Ref-Tx) from 18,557 human genes.
• Compared identified Ref-Tx with MANE-select and APPRIS annotations.

Main Results:

• Identified 18,094 dominant Ref-Tx from 18,557 genes using long-read data.
• 14,546 Ref-Tx matched MANE-select transcripts, indicating high concordance.
• The long-read dataset showed better agreement with MANE genes and more topmost Ref-Tx compared to potential confounding factors.

Conclusions:

• Long-read sequencing provides a more accurate assessment of dominant protein-coding transcripts than short-read data.
• The identified Ref-Tx and the developed eCPG tool enhance the understanding of tissue-specific gene expression.
• This work facilitates the accurate identification of bona fide translated peptide products.