Boosting Gene Mapping Power and Efficiency with Efficient Exact Variance Component Tests of Single Nucleotide...

Area of Science:

• Genetics and Genomics
• Statistical Bioinformatics
• Computational Biology

Background:

• Single nucleotide polymorphism (SNP) set tests are crucial for analyzing next-generation sequencing (NGS) data.
• Existing methods like the Sequence Kernel Association Test (SKAT) rely on asymptotic theory, leading to conservatism and reduced power with small or moderate sample sizes.
• Limited sample sizes in current NGS studies necessitate more powerful and accurate statistical approaches.

Purpose of the Study:

• To derive and implement computationally efficient, exact (nonasymptotic) statistical tests for analyzing SNP sets in NGS data.
• To evaluate the performance of these novel tests, termed ExactVCTest (including eScore, eLRT, and eRLRT), particularly in scenarios with limited sample sizes.
• To compare the power and type I error control of ExactVCTest against established methods like SKAT and SKAT-optimal (SKAT-o).

Main Methods:

• Derivation of exact score (eScore), likelihood ratio (eLRT), and restricted likelihood ratio (eRLRT) tests.
• Implementation of these tests within a software package using the Julia programming language.
• Simulation studies across various genetic scenarios and sample sizes to assess performance.
• Application of the developed tests to a real-world exome sequencing dataset.

Main Results:

• ExactVCTest (eScore, eLRT, eRLRT) demonstrated well-controlled type I error rates across simulations.
• eScore P-values were consistently smaller than those from SKAT under the alternative model.
• eLRT and eRLRT exhibited significantly higher statistical power compared to eScore, SKAT, and SKAT-o across diverse scenarios and sample sizes.
• Application to exome sequencing data successfully replicated previous findings and provided insights into rare variant effects within genes.

Conclusions:

• The developed ExactVCTest provides a computationally efficient and statistically powerful alternative for SNP set association testing in NGS data, especially beneficial for small sample sizes.
• eLRT and eRLRT tests are particularly recommended for their superior power in detecting genetic associations.
• The freely available Julia software package facilitates the application of these advanced statistical methods in genetic research.