inDelphi

inDelphi predicts DNA repair outcomes at CRISPR/SpCas9-induced double-strand breaks, forecasting genotypes and frequencies of non-homologous end-joining (NHEJ)-mediated insertions and deletions to enable precise genome-editing applications.

Key Features:

• Predictive modeling: Uses machine learning to predict genotypes and frequencies of deletions from 1 to 60 base pairs and 1 base-pair insertions at Cas9-cleaved sites.
• Quantitative accuracy: Achieves a correlation of r = 0.87 in outcome frequency predictions across five different human and mouse cell lines.
• Precise-50 guide RNA identification: Identifies guide RNAs that yield a single predominant genotype comprising ≥50% of major editing products, with ~5–11% of human-targeting guides falling into this category.

Scientific Applications:

• Disease mutation correction: Validated correction of pathogenic alleles in primary patient-derived fibroblasts for Hermansky-Pudlak syndrome and Menkes disease.
• Genome-editing precision without donor templates: Predicts template-free Cas9 editing outcomes to support precise genetic corrections beyond gene disruption.

Methodology:

A training dataset comprising a library of 2,000 Cas9 guide RNAs paired with DNA target sites was constructed and used to train the model; the tool employs machine learning algorithms to analyze sequence patterns and predict NHEJ repair outcomes.