Identifying rare-variant associations in parent-child trios using a Gaussian support vector machine.

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• High-throughput sequencing enables focus on rare variants (RVs) for complex traits.
• Analyzing single RVs lacks statistical power, necessitating new approaches.
• Parent-child trios offer a framework for studying variant transmission.

Purpose of the Study:

• Develop and apply a novel statistical approach, Trio-SVM, to aggregate and evaluate RV transmissions in trios.
• Assess the performance of Trio-SVM in detecting gene associations with complex traits.
• Investigate the impact of sample size on the power of Trio-SVM.

Main Methods:

• Calculated an initial score based on transmission distortion of RVs from parents to offspring.
• Utilized a support vector machine (SVM) to nonlinearly map transmission distortion scores.
• Applied Trio-SVM to 275 trios from the Genetic Analysis Workshop 18 (GAW18) data.
• Evaluated type I error rates and statistical power using simulated trait values.

Main Results:

• Trio-SVM demonstrated appropriate type I error control with simulated data.
• The method lacked sufficient power with a sample size of 267 trios.
• Larger sample sizes (500-1000 trios) provided adequate power.
• Two candidate genes on chromosome 3 showed marginal associations with hypertension in real GAW18 data.

Conclusions:

• Trio-SVM is a viable approach for analyzing rare variants in parent-child trios.
• The method's power is dependent on sample size, requiring larger cohorts for robust findings.
• Further validation in larger datasets is warranted to confirm associations with complex traits like hypertension.