Transformers significantly improve splice site prediction
Benedikt A. Jónsson, Gísli H. Halldórsson, Steinþór Árdal, Sölvi Rögnvaldsson, Eyþór Einarsson, Patrick Sulem, Daníel F. Guðbjartsson, Páll Melsted, Kári Stefánsson, Magnús Ö. Úlfarsson · 2024
Read Transformers significantly improve splice site prediction by Benedikt A. Jónsson, Gísli H. Halldórsson, Steinþór Árdal, Sölvi Rögnvaldsson, Eyþór Einarsson, Patrick Sulem, Daníel F. Guðbjartsson, Páll Melsted, Kári Stefánsson, Magnús Ö. Úlfarsson (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-11618611.txt.
Description
Benedikt Jónsson and nine co-authors built a transformer model that finds splice sites in DNA far more accurately than the standard tool. Splice sites mark where gene messages are cut and rejoined, and errors there cause disease. Their model reads raw sequences up to forty-five thousand letters long, compresses them with a neural network, then focuses attention on candidate sites.
Against the leading program, SpliceAI, the new method won on several fronts. It matched gene catalogs better and found more real junctions in RNA data from nearly eighteen thousand Icelanders and a large tissue project. It also ranked disease-linked variants in a clinical database slightly higher. The gains are modest in number but broad in scope. The authors argue better splicing prediction will sharpen both genetic research and diagnosis.
Catalog record
| CREATOR | Benedikt A. Jónsson, Gísli H. Halldórsson, Steinþór Árdal, Sölvi Rögnvaldsson, Eyþór Einarsson, Patrick Sulem, Daníel F. Guðbjartsson, Páll Melsted, Kári Stefánsson, Magnús Ö. Úlfarsson |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |