United StatesHealthcare and life sciences2025 – ongoingScribe Therapeutics
DeepXE model for predicting guide RNA editing efficiency
Scribe Therapeutics said its in-house DeepXE model doubled the hit rate and cut guide RNA screening by more than half
Problem
Designing potent guide RNAs for the company's CasXE gene editors normally needs large screening runs in the lab. Scribe wanted to speed up the design of potent editing molecules.
Approach
Scribe built DeepXE, a proprietary machine learning model that predicts the on-target editing efficiency of guide RNAs for its XE variant of CRISPR-CasX. The company unveiled it in September 2025 at a Cold Spring Harbor Laboratory genome engineering meeting. It said it tested the model across several engineered XE variants, target genes and assay conditions.
Results
Takeaway
The results are company-reported and preclinical. Scribe gave no training data or independent validation.
Categories
- Size
- Small business
- Industry
- Healthcare and life sciences
- Business function
- IT and R&D
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Scribe's release introduced DeepXE, which it called the first machine learning model able to accurately predict on-target editing efficiency for XE. It reported a 2x higher hit rate, under 10% false negatives, sensitivity above 90% and a screening size cut of more than 50%. It said the model beat conventional Cas9-based models.
Full details are in the original sources.