Structure-based network analysis predicts pathogenic variants in human proteins associated with inherited retinal...

Area of Science:

• Genomics and Bioinformatics
• Structural Biology
• Ophthalmology

Background:

• Gene sequencing identifies variants, but tools to determine pathogenicity are lacking.
• Inherited retinal disease (IRD) gene therapy requires precise identification of disease-causing variants.
• Existing missense prediction tools often rely on specific outcome data.

Purpose of the Study:

• To apply Structure-Based Network Analysis (SBNA) for identifying residues prone to disease-causing missense mutations in IRD.
• To validate SBNA's efficacy using established disease proteins and deep mutagenesis data.
• To discover novel causative variants in IRD patients with unidentified genetic causes.

Main Methods:

• Structure-based network analysis (SBNA) based on structural first principles.
• Correlation analysis with deep mutagenesis data for BRCA1, HRAS, PTEN, and ERK2.
• Application to 47 IRD genes with crystal structure data, validated against ClinVar.
• Machine learning models incorporating SBNA, BLOSUM62, and EVE scores for variant pathogenicity prediction.

Main Results:

• SBNA scores strongly correlated with deep mutagenesis data in validation proteins.
• SBNA reliably identified disease-causing variants in 47 IRD genes.
• A predictive model using SBNA achieved an AUC of 0.851 for novel IRD variants.
• Incorporating EVE scores improved model performance to an AUC of 0.927.
• Likely causative variants were identified in 40 of 63 IRD patients.

Conclusions:

• SBNA is a robust, unbiased method for identifying disease-causing variants in human proteins.
• SBNA can aid in predicting variants causative of IRD, facilitating genetic diagnosis.
• This approach holds promise for advancing gene therapy targets in inherited retinal diseases.