Accurately modeling RNase H-mediated antisense oligonucleotide efficacy

Area of Science:

• Biotechnology
• Genomics
• Drug Discovery

Background:

• Antisense oligonucleotides (ASOs) show therapeutic potential but predicting their efficacy is challenging.
• Current methods require extensive and costly experimental screening for optimal ASO candidate identification.

Purpose of the Study:

• To develop a predictive model for ASO efficacy using a comprehensive dataset.
• To create a systematic resource for evaluating ASO design parameters and accelerating drug development.

Main Methods:

• Compiled ASO Atlas, a database of 188,521 RNase H-mediated ASO sequences and efficacy data.
• Trained OligoAI, a deep learning model, to predict in vitro efficacy by integrating RNA target context, ASO sequence, chemistry, and dosage.
• Experimentally validated OligoAI performance by targeting KCNT2.

Main Results:

• OligoAI demonstrated accurate prediction of ASO efficacy.
• Experimental validation showed a 5.72-fold reduction in screening effort compared to random selection.
• ASO Atlas serves as a resource for evaluating ASO design hypotheses.

Conclusions:

• ASO Atlas and OligoAI provide a powerful, accessible platform for optimizing ASO design.
• The developed tools facilitate accelerated development of ASO-based therapeutics.
• This approach significantly reduces the experimental burden in identifying effective ASO drug candidates.