Structure-guided isoform identification for the human transcriptome

• DOI: 10.7554/elife.82556
• Publication Date: 2022-12-15T11:00:45Z
• Authors: Markus J Sommer; Sooyoung Cha; Ales Varabyou; Natalia Rincon; Sukhwan Park; Ilia Minkin; Mihaela Pertea; Martin Steinegger; Steven L Salzberg
• Keywords: Human proteome project; Proteome; Gene Annotation; Gene prediction; Identification

Recently developed methods to predict three-dimensional protein structure with high accuracy have opened new avenues for genome and proteome research. We explore a hypothesis in annotation, namely whether computationally predicted structures can help identify which of multiple possible gene isoforms represents functional product. Guided by predictions, we evaluated over 230,000 human protein-coding genes assembled from 10,000 RNA sequencing experiments across many tissues. From this set transcripts, identified hundreds more confidently potentially superior function comparison canonical the latest database. illustrate our method examples where provides guide combination expression evolutionary evidence. Additionally, provide complete as resource better understand their isoforms. These results demonstrate promise prediction annotation tool, allowing us refine even most highly curated catalog proteins. More generally practical, structure-guided approach that be used enhance any genome.