Joint estimation and imputation of variant functional effects using high throughput assay data.

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Deep mutational scanning (DMS) assays offer high-throughput variant functional assessment.
• DMS phenotypic data correlate with clinical outcomes but suffer from individual variant-level noise.
• Accurate variant impact prediction is crucial for genetic disease diagnosis and research.

Approach:

• Developed a novel framework, Functional Substitution Estimation (FUSE), to jointly estimate variant impact.
• Leveraged related measurements within and across experimental assays for robust estimation.
• Integrated a large corpus of DMS data, estimating mean functional effects per residue and normalizing by substitution type.

Key Points:

• FUSE enhances correlation between functional screening datasets for identical variants.
• Significantly improved separation of pathogenic and benign variants (ClinVar BRCA1, p=2.24×10⁻⁵¹).
• Expanded variant prediction from 2,741 to 10,347 by inferring unscreened substitution effects.
• Demonstrated improved cancer syndrome prediction in UK Biobank TP53 variant carriers (p=1.77×10⁻⁶).

Conclusions:

• The FUSE framework offers improved estimates of variant impact.
• Broadens the utility of data generated from functional screening assays.
• Promises to enhance genetic variant interpretation and clinical applications.