DeepRNAi: Predicting on- and off-target siRNA activity and establishing a benchmark

DeepRNAi predicts siRNA on-target and off-target knockdown from siRNA and mRNA sequence with experimental-level accuracy, trained on 16 million siRNA–transcript pairs.

• Why it matters: Off-target silencing can create toxicity and regulatory hurdles, yet qPCR and RNA-seq assessment is slow and constrained by limited, species-appropriate cell models.
• What they did: DeepRNAi was trained on 16 million siRNA–transcript pairs from 2,031 siRNAs, using transcript context from biological foundation models to model activity.
• The result: It generalizes across cell models, transcripts, genomes, and species, distinguishes approved therapeutics from hepatotoxic candidates, and a survey of 24.4 million siRNAs shows 90.1% of the 100 most potent have off-target concerns.

AI summary of the abstract · may contain errors · not medical advice · read the paper.