EukaUTR: a foundation model unifying functional modelling and design of eukaryotic 3′ UTRs
TL;DR
Across 13 prediction tasks spanning post-transcriptional regulation, RNA fate and expression output, EukaUTR models matched or exceeded the strongest external baselines on nearly all tasks, with relative improvements of up to 27.45%.
Abstract
Eukaryotic mRNA 3′ UTRs encode regulatory information that shapes post-transcriptional control, RNA fate and gene expression. However, a 3′ UTR-specific foundation model that spans broad eukaryotic sequence diversity while supporting both functional prediction and sequence design is lacking. Here we present EukaUTR, a 3′ UTR-specific foundation model trained on a large-scale eukaryotic 3′ UTR sequence corpus spanning diverse evolutionary lineages. Across 13 prediction tasks spanning post-transcriptional regulation, RNA fate and expression output, EukaUTR models matched or exceeded the strongest external baselines on nearly all tasks, with relative improvements of up to 27.45%. EukaUTR also generated de novo 3′ UTRs with natural-like sequence and regulatory properties. EukaUTR-Guide further derived stability-associated signals from small sequence sets to guide editing towards enhanced predicted stability. Together, EukaUTR provides a sequence-to-function-to-design framework for transferable 3′ UTR modelling and design.
