Decoding the interactions and functions of non-coding RNA with artificial intelligence

In addition to encoding proteins, mRNAs have context-specific regulatory roles that contribute to many cellular processes. However, uncovering new mRNA functions is constrained by limitations of traditional biochemical and computational methods. In this Roadmap, we highlight how artificial intellige...

Full description

Saved in:
Bibliographic Details
Published in:Nature reviews. Molecular cell biology Vol. 26; no. 10; p. 797
Main Authors: Jung, Vincent, Vincent-Cuaz, Cédric, Tumescheit, Charlotte, Fournier, Lisa, Darsinou, Marousa, Xu, Zhi Ming, Saadat, Ali, Wang, Yiran, Tsantoulis, Petros, Michielin, Olivier, Fellay, Jacques, Patani, Rickie, Ramos, Andres, Frossard, Pascal, Hastings, Janna, Riccio, Antonella, van der Plas, Lonneke, Luisier, Raphaëlle
Format: Journal Article
Language:English
Published: England 01.10.2025
Subjects:
Animals
Artificial Intelligence
Humans
Large Language Models
RNA, Messenger - genetics
RNA, Messenger - metabolism
RNA, Untranslated - genetics
RNA, Untranslated - metabolism
Sequence Analysis, RNA - methods
ISSN:1471-0080, 1471-0080
Online Access:Get more information
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In addition to encoding proteins, mRNAs have context-specific regulatory roles that contribute to many cellular processes. However, uncovering new mRNA functions is constrained by limitations of traditional biochemical and computational methods. In this Roadmap, we highlight how artificial intelligence can transform our understanding of RNA biology by fostering collaborations between RNA biologists and computational scientists to drive innovation in this fundamental field of research. We discuss how non-coding regions of the mRNA, including introns and 5' and 3' untranslated regions, regulate the metabolism and interactomes of mRNA, and the current challenges in characterizing these regions. We further discuss large language models, which can be used to learn biologically meaningful RNA sequence representations. We also provide a detailed roadmap for integrating large language models with graph neural networks to harness publicly available sequencing and knowledge data. Adopting this roadmap will allow us to predict RNA interactions with diverse molecules and the modelling of context-specific mRNA interactomes.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
ObjectType-Review-3
content type line 23
ISSN:1471-0080
1471-0080
DOI:10.1038/s41580-025-00857-w