A New Efficient Method to Detect Genetic Interactions for Lung Cancer GWAS

Jennifer Luyapan, Xuemei Ji, Siting Li, Xiangjun Xiao, Dakai Zhu, Eric J. Duell, David C. Christiani, Matthew B. Schabath, Susanne M. Arnold, Shanbeh Zienolddiny, Hans Brunnström, Olle Melander, Mark Thornquist, Todd A. MacKenzie, Christopher I. Amos, Jiang Gui

Abstract

Genome-wide association studies (GWAS) have proven successful in predicting genetic risk of disease using single-locus models; however, identifying single nucleotide polymorphism (SNP) interactions at the genome-wide scale is limited due to computational and statistical challenges. We addressed the computational burden encountered when detecting SNP interactions for survival analysis, such as age of disease-onset. To confront this problem, we developed a novel algorithm, called the Efficient Survival Multifactor Dimensionality Reduction (ES-MDR) method, which used Martingale Residuals as the outcome parameter to estimate survival outcomes, and implemented the Quantitative Multifactor Dimensionality Reduction method to identify significant interactions associated with age of disease-onset. To demonstrate efficacy, we evaluated this method on two simulation data sets to estimate the type I error rate and power. Simulations showed that ES-MDR identified interactions using less computational workload and allowed for adjustment of covariates. We applied ES-MDR on the OncoArray-TRICL Consortium data with 14,935 cases and 12,787 controls for lung cancer (SNPs = 108,254) to search over all two-way interactions to identify genetic interactions associated with lung cancer age-of-onset. Our experiment identified many one-way and two-way models with a single-base deletion in the noncoding region of BRCA1 (HR = 1.24, P = 3.15 x 10⁻¹⁵) as the top marker to predict age of lung cancer onset. From the results of our extensive simulations and analysis of a large GWAS study, we demonstrated that our method is an efficient algorithm that identified genetic interactions to include in our models to predict survival outcomes.

Introduction

While traditional Genome-Wide Association Studies (GWAS) have been instrumental in identifying single genetic variants (SNPs) linked to disease, they often fail to capture the complexity of polygenic conditions. Many complex diseases, including lung cancer, are influenced by epistasis—the interaction between multiple genes. Identifying these interactions is a significant challenge, particularly for survival analysis where the goal is to predict the age of disease onset rather than a binary diagnosis.

Existing methods, such as Surv-MDR, rely on log-rank tests that are computationally expensive and struggle to account for confounding factors like smoking or age. This creates a "needle in a haystack" problem where the computational burden of testing millions of SNP-SNP combinations becomes prohibitive. The ES-MDR method was developed to provide a statistically robust, covariate-adjusted, and computationally efficient alternative.

Methods

The core innovation of the Efficient Survival Multifactor Dimensionality Reduction (ES-MDR) method lies in its two-stage approach to survival data:

  1. The Martingale Residual Transformation: Instead of processing raw survival times and event statuses, which require complex iterative calculations, the method calculates Martingale residuals from a Cox proportional hazards model. This allows researchers to incorporate covariate adjustments (e.g., smoking status) directly into a single continuous score for each patient.
The Martingale residual calculation incorporates the hazard function, SNP data, and covariates into a single score.
  1. Quantitative MDR (QMDR): The algorithm then applies QMDR to these residuals. Genotype combinations are categorized into "high-risk" or "low-risk" groups based on the sign of the residual sum. By transforming a survival problem into a continuous variable problem, ES-MDR can utilize a cross-validated t-test to identify optimal interaction models, bypassing slow survival calculations.

Results and Performance

The authors validated ES-MDR using both simulated datasets and real-world genomic data from the OncoArray-TRICL Consortium (14,935 cases and 12,787 controls).

Computational Efficiency and Power

The most significant advantage of ES-MDR is its speed. In testing, ES-MDR processed 100 simulated datasets in just 2.25 minutes, whereas the previous Surv-MDR method required 734.5 minutes—a speedup of approximately 326x. Furthermore, simulation results indicated that ES-MDR maintained or improved statistical power across various heritability and sample size scenarios.

Table 2 illustrates the percent change in power between ES-MDR and Surv-MDR across different genetic architectures and sample sizes.

Biological Findings in Lung Cancer

When applied to the OncoArray-TRICL data (analyzing 108,254 SNPs), the method identified a single-base deletion in the noncoding region of BRCA1 (rs749410065) as the primary predictor for the age of lung cancer onset (Hazard Ratio = 1.24). The study also uncovered significant two-way interactions between BRCA1 and other genes such as CBR1 and LOC100133286. By adjusting for smoking status within the residuals, the researchers ensured that these genetic findings were independent of the primary environmental driver of the disease.

Discussion and Conclusion

ES-MDR represents a significant step forward in genomic survival analysis. By utilizing Martingale residuals as a mathematical surrogate for log-rank statistics, the method achieves massive efficiency gains without sacrificing the ability to adjust for critical covariates.

However, some limitations remain. The use of residuals may result in minor information loss compared to raw survival data, and the model showed decreased accuracy for early-onset cases (under age 40) due to data sparsity. Future research will focus on incorporating biological pathway information to further refine SNP selection and exploring higher-order (3-way and 4-way) interactions. Overall, ES-MDR provides a scalable framework for discovering the complex genetic architectures underlying disease progression.

Summary updated

Abstract

Abstract Background: Genome-wide association studies (GWAS) have proven successful in predicting genetic risk of disease using single-locus models; however, identifying single nucleotide polymorphism (SNP) interactions at the genome-wide scale is limited due to computational and statistical challenges. We addressed the computational burden encountered when detecting SNP interactions for survival analysis, such as age of disease-onset. To confront this problem, we developed a novel algorithm, called the Efficient Survival Multifactor Dimensionality Reduction (ES-MDR) method, which used Martingale Residuals as the outcome parameter to estimate survival outcomes, and implemented the Quantitative Multifactor Dimensionality Reduction method to identify significant interactions associated with age of disease-onset. Methods: To demonstrate efficacy, we evaluated this method on two simulation data sets to estimate the type I error rate and power. Simulations showed that ES-MDR identified interactions using less computational workload and allowed for adjustment of covariates. We applied ES-MDR on the OncoArray-TRICL Consortium data with 14,935 cases and 12,787 controls for lung cancer (SNPs = 108,254) to search over all two-way interactions to identify genetic interactions associated with lung cancer age-of-onset. We tested the best model in an independent data set from the OncoArray-TRICL data. Results: Our experiment on the OncoArray-TRICL data identified many one-way and two-way models with a single-base deletion in the noncoding region of BRCA1 (HR = 1.24, P = 3.15 x 10 -15 ), as the top marker to predict age of lung cancer onset. Conclusions: From the results of our extensive simulations and analysis of a large GWAS study, we demonstrated that our method is an efficient algorithm that identified genetic interactions to include in our models to predict survival outcomes.
BibTeX
@inproceedings{luyapan2020new,
  author = {Jennifer Luyapan and Xuemei Ji and Siting Li and Xiangjun Xiao and Dakai Zhu and Eric J. Duell and David C. Christiani and Matthew B. Schabath and Susanne M. Arnold and Shanbeh Zienolddiny and Hans Brunnström and Olle Melander and Mark Thornquist and Todd A. MacKenzie and Christopher I. Amos and Jiang Gui},
  title = {{A New Efficient Method to Detect Genetic Interactions for Lung Cancer GWAS}},
  booktitle = {Research},
  year = {2020},
  doi = {10.21203/rs.2.14850/v3},
}
Figure