Leveraging a Small Language Model to Achieve Full-Length mRNA Analysis with Improved Prediction of Stability & Translation Efficiency
Time: 12:25 pm
day: Day One
Details:
- CodonBERT is pre-trained on 10 million mRNA sequences to capture codon dependencies and interactions across different species
- mRNA-LM extends CodonBERT from the CDS to full-length mRNA by integrating two pre-trained BERT models on UTR regions
- Both CodonBERT and mRNA-LM show significant improvements in predicting mRNA properties compared to existing methods
