TRIO RVEMVS: A Bayesian framework for rare variant association analysis with expectation-maximization variable...

Area of Science:

• Genetics
• Statistical Genetics
• Computational Biology

Background:

• Rare variants are increasingly recognized for their role in complex diseases.
• Individual rare variant association testing is challenging due to low allele frequencies.
• Existing methods often struggle to detect associations at the individual variant level.

Purpose of the Study:

• To develop and evaluate a novel method, TRIO_RVEMVS, for simultaneous detection of common and rare variants.
• To assess the performance of TRIO_RVEMVS at both the individual variant and regional levels.
• To apply the method to real-world data for identifying disease-associated variants.

Main Methods:

• Developed an expectation maximization variable selection (EMVS) method for family trio data.
• Simulated large (1500 families) and small (350 families) datasets to test TRIO_RVEMVS.
• Compared TRIO_RVEMVS performance against PedGene and RV-TDT using various analytical approaches.

Main Results:

• TRIO_RVEMVS outperformed existing methods at the region level when common variants were included.
• TRIO_RVEMVS showed competitive performance with PedGene and outperformed RV-TDT for rare variant analysis.
• In simulations, TRIO_RVEMVS achieved true positive rates of 12.20-13.10% and false positive rates of 0.74-1.30% for individual rare variants.
• Applied to Kids First data, TRIO_RVEMVS identified 3 rare variants associated with orofacial clefts.

Conclusions:

• TRIO_RVEMVS is an effective method for detecting both common and rare variants at the individual level using family trios.
• The method demonstrates improved performance over existing approaches, particularly for rare variant association.
• TRIO_RVEMVS successfully identified potential risk variants for orofacial clefts in a real-world dataset.