Fast computation of kernel statistics using genotype value decomposition

TL;DR

This paper introduces a genotype value decomposition method that significantly accelerates kernel-based genetic association tests like SKAT, enabling efficient analysis of large-scale human genetic data.

Contribution

The paper presents a novel genotype value decomposition approach that reduces SKAT computation time from quadratic to linear complexity, facilitating large-scale genetic studies.

Findings

Kernel matrix can be derived from genotype value vectors.

The method reduces SKAT computation time to O(n).

Enables efficient analysis of large human genetic datasets.

Abstract

Because of the recent advances of genome sequences, a large number of human genome sequences are available for the study of human genetics. Genome-wide association studies typically focus on associations between single-nucleotide polymorphisms and traits such as major human diseases. However, the statistical power of classical single-marker association analysis for rare variants is limited. To address the challenge, rare and low-frequency variants are often grouped into a gene or pathway level, and the effects of multiple variants evaluated based on collapsing methods. The sequential kernel association test (SKAT) is one of the most effective collapsing methods. SKAT utilizes the kernel matrix. The size of the kernel matrix is O(n^2), where the sample size is n, so that the calculation of the data using the kernel method requires a long time. As the sample sizes of human genetic studies increase, the computational time is getting more and more problematic. In the present paper, the genotype value decomposition method is proposed for the handling the sequential kernel in a short period of time. The method can be referred to as genotype value decomposition. In the present paper, it is shown that the genetic relationship matrix and Identity by State (IBS) matrix can be obtained using the genotype value vectors. By using this method, the SKAT can be conducted with time complexity O(n). The proposed method enables to conduct SKAT for samples of human genetics.