CAVaLRi: An Algorithm for Rapid Identification of Diagnostic Germline Variation

Area of Science:

• Genomics
• Medical Genetics
• Bioinformatics

Background:

• Clinical exome and genome sequencing (ES/GS) are vital for diagnosing rare genetic diseases (RGD).
• Interpreting the large volume of variants from ES/GS poses a significant challenge in clinical practice.
• Existing gene prioritization tools struggle with the scale and complexity of variant interpretation.

Purpose of the Study:

• To introduce Clinical Assessment of Variants by Likelihood Ratios (CAVaLRi), a novel framework for prioritizing diagnostic genes in RGD.
• To evaluate CAVaLRi's performance against leading gene prioritization algorithms using clinical ES data.
• To assess CAVaLRi's robustness with both clinician-curated and computationally derived phenotype data.

Main Methods:

• Developed CAVaLRi, a modified likelihood ratio (LR) framework incorporating variant impact, genotype, phenotype, and segregation data.
• Trained and tested CAVaLRi on an internal cohort of 655 clinical ES cases.
• Validated CAVaLRi against Exomiser (hiPHIVE, PhenIX), LIRICAL, and XRare using a distinct cohort of 12,832 ES cases.

Main Results:

• CAVaLRi significantly outperformed existing algorithms in precision and diagnostic gene rank, particularly with clinician-curated phenotypes (PR AUC: 0.701, Rank: 1.59).
• CAVaLRi maintained high precision with computationally derived phenotypes (PR AUC: 0.658, Rank: 1.68), outperforming other tools.
• In a large, heterogeneous cohort, CAVaLRi demonstrated superior precision (PR AUC: 0.335, Rank: 1.91).

Conclusions:

• CAVaLRi offers a robust and scalable solution for prioritizing diagnostic genes in rare genetic diseases.
• The algorithm effectively handles noisy, computationally derived phenotypes, improving diagnostic efficiency.
• CAVaLRi addresses the growing demand for accurate ES/GS interpretation by focusing on the most relevant variants.