A Comparative Study of Joint-SNVs Analysis Methods and Detection of Susceptibility Genes for Gastric Cancer in Korean Population

Jin-Xiong Lv, Shikui Tu, Lei XuIScIDE 2017 · OpenReview

Abstract

Many joint-SNVs (single-nucleotide variants) analysis methods were proposed to tackle the "missing heritability" problem, which emphasizes that the joint genetic variants can explain more heritability of traits and diseases. However, there is still lack of a systematic comparison and investigation on the relative strengths and weaknesses of these methods. In this paper, we evaluated their performance on extensive simulated data generated by varying sample size, linkage disequilibrium (LD), odds ratios (OR), and minor allele frequency (MAF), which aims to cover almost all scenarios encountered in practical applications. Results indicated that a method called Statistics-space Boundary Based Test (S-space BBT) showed stronger detection power than other methods. Results on a real dataset of gastric cancer for Korean population also validate the effectiveness of the S-space BBT method.

Introduction

Genome-Wide Association Studies (GWAS) have been instrumental in identifying common genetic variants associated with complex diseases. However, these variants often account for only a small portion of phenotypic heritability. This "missing heritability" is hypothesized to reside in rare variants (MAF 1-5%) and the combined effects of multiple SNVs. Joint-SNV analysis methods aim to capture these effects by analyzing groups of variants simultaneously. This study provides a systematic evaluation of six such methods, highlighting the strengths of the Statistics-space Boundary Based Test (S-space BBT).

Methodology

The researchers compared six representative joint-SNV analysis methods categorized by their statistical approach:

  • Projection Methods: These include Hotelling’s T Square Test, the Sumstat Test (Burden Test), and regression-based methods like SKAT and SKATO.
  • Probability Space Combination: Fisher’s Combined Test, which aggregates individual P-values.
  • High-Dimensional Multivariate Testing: The S-space BBT method, which performs testing directly in high-dimensional statistic space using a boundary-based rejection domain.

S-space BBT is distinguished by its ability to account for both the direction and the correlation of genetic components without the information loss typical of averaging or simple summation used in burden tests.

Simulation Results and Comparative Analysis

The performance of these methods was evaluated using extensive simulated data (Datasets 1-5) generated via PLINK, varying sample sizes (100 to 1000 case/control pairs), linkage disequilibrium (LD), odds ratios (OR), and minor allele frequencies (MAF).

Detection power of six methods on Dataset 1 across different sample sizes, showing the consistent lead of S-space BBT.

Across all simulated scenarios, S-space BBT consistently demonstrated superior detection power. A critical finding was the method's performance with rare variants (MAF = 1%). While traditional methods often saw power drop below 20%, S-space BBT maintained a power exceeding 60%, supporting its utility for the "Common Disease, Rare Variant" (CDRV) hypothesis.

Comparative detection power across Datasets 3, 4, and 5, demonstrating the consistent superiority of S-space BBT (darkest bar) as sample size increases.

As shown in the figures, the power of all methods generally increases with sample size, but S-space BBT maintains a significant lead, particularly in more complex genetic architectures where other tests lose sensitivity.

Real-World Validation: Gastric Cancer and Bipolar Disorder

The effectiveness of S-space BBT was further validated using empirical datasets. In a Korean gastric cancer dataset (GSE58356), the method identified the PSCA gene with high significance (P = 8.90E-10). Furthermore, 65% of the top 20 identified susceptibility genes were independently verified as cancer-related through literature review.

In a secondary benchmark using a bipolar disorder dataset (GSE71443), S-space BBT successfully identified known susceptibility loci (ANK3 and PALB2) with higher sensitivity than competing methods, proving its effectiveness even in smaller populations.

Conclusion

The S-space BBT method offers a robust and powerful alternative for joint-SNV analysis. By effectively handling rare variants and maintaining signal integrity in high-dimensional spaces, it addresses key challenges in identifying the genetic basis of complex diseases. Future work will focus on addressing sample disequilibrium and optimizing computational efficiency for larger-scale applications.

Summary updated

Abstract

Many joint-SNVs (single-nucleotide variants) analysis methods were proposed to tackle the "missing heritability" problem, which emphasizes that the joint genetic variants can explain more heritability of traits and diseases. However, there is still lack of a systematic comparison and investigation on the relative strengths and weaknesses of these methods. In this paper, we evaluated their performance on extensive simulated data generated by varying sample size, linkage disequilibrium (LD), odds ratios (OR), and minor allele frequency (MAF), which aims to cover almost all scenarios encountered in practical applications. Results indicated that a method called Statistics-space Boundary Based Test (S-space BBT) showed stronger detection power than other methods. Results on a real dataset of gastric cancer for Korean population also validate the effectiveness of the S-space BBT method.
BibTeX
@inproceedings{lv2017comparative,
  author = {Jin-Xiong Lv and Shikui Tu and Lei Xu},
  title = {{A Comparative Study of Joint-SNVs Analysis Methods and Detection of Susceptibility Genes for Gastric Cancer in Korean Population}},
  booktitle = {IScIDE},
  year = {2017},
  url = {https://openreview.net/forum?id=7C7G30JUFkW},
}
Figure