UTR-Insight: integrating deep learning for efficient 5' UTR discovery and design

Area of Science:

• Biotechnology
• Computational Biology
• Genomics

Background:

• The 5' untranslated region (UTR) significantly influences mRNA stability and translation efficiency, crucial factors for therapeutic applications.
• Current predictive models for 5' UTR function often lack the accuracy needed for high-throughput screening in drug development.

Purpose of the Study:

• To develop a highly accurate computational model, UTR-Insight, for predicting the impact of 5' UTR sequences on protein expression.
• To utilize UTR-Insight for large-scale in silico screening to identify novel 5' UTR sequences with enhanced translational capabilities.

Main Methods:

• Development of UTR-Insight, a hybrid model combining a pretrained language model with a CNN-Transformer architecture.
• Training and validation of UTR-Insight on both random and endogenous 5' UTR sequences to assess its predictive power for mean ribosome load (MRL).
• High-throughput in silico screening of hundreds of thousands of endogenous 5' UTRs from diverse species (primates, mice, viruses) using UTR-Insight.

Main Results:

• UTR-Insight achieved high accuracy, explaining 89.1% of MRL variation in random 5' UTRs and 82.8% in endogenous 5' UTRs, outperforming existing models.
• In silico screening identified numerous endogenous 5' UTRs that increased protein expression up to 319% compared to the standard human α-globin 5' UTR.
• UTR-Insight-designed sequences demonstrated even higher protein expression levels than the top-performing endogenous sequences found during screening.

Conclusions:

• UTR-Insight is a powerful and accurate tool for predicting 5' UTR function and optimizing protein expression.
• The study identified novel 5' UTR sequences with significant potential for enhancing therapeutic protein production.
• This approach accelerates the discovery of regulatory elements for improved gene and protein expression in biotechnology and medicine.