img

Title: Developing a Genetic Risk Index for Peanut Allergy

Denise Daley

University of British Columbia, Canada

Biography

Dr.Daley completed a PhD in Epidemiology and Biostatistics at Case Western Reserve University in 2003, followed by post-doctoral training at the University of British Columbia from 2003-2008. In 2008 she was awarded a Tier II Canadian Research Chair (genetic epidemiology of common complex diseases), renewed in 2013 and she is currently an Associate Professor in the Faculty of Medicine at the University of British Columbia. Dr. Daley’s interests are in the study of complex diseases such as asthma, food allergy, allergic disease, cancer and heart disease, with a focus on gene-gene and gene-environment interactions.

Dr. Daley is studying genetic susceptibility to asthma and other allergic conditions and the complex epigenetic mechanisms that may be involved. She is working to determine what contribution gender, genes, and environment make to the development of asthma and how the epigenome responds to environmental exposures such as tobacco smoke.

Abstract

Over 200 single nucleotide polymorphisms (SNPs) have been found to be associated with food allergy (FA) in genome-wide association studies (GWAS). A Genetic risk score (GRS),  is an index that can be derived from genome-wide association studies to summarize the genetic risk encompassed by a set of SNPs, and is useful in risk stratification and prediction. Our objective was to use information from the Canadian Peanut Allergy Registry (CanPAR) GWAS study [1] to develop a GRS and evaluate the positive predictive value of the GRS in CanPAR and the Canadian Asthma Primary Prevention Study (CAPPS).

Methods

Our study aims to use the food allergy (FA)-associated SNPs using p-value thresholds ranging from 1.0 * 10-4 to 1.0 * 10-6 to generate a GRS using a weighted sum of the number of risk alleles (with values 0/1/2). Weighting each SNP by the natural log of their respective odds ratio (OR). We then evaluated the area under the curve (AUC) which is used to determine the effectiveness of the classification and the positive predictive value (PPV). The AUC value ranges from .5 to 1 with .5 being a poor classifier and 1 a perfect fit.

Results

Table 1 Summary of GRS risk model by three different p-value thresholds

 

# of SNPs selected

Statistical measures

p-value Threshold

Genotyped

Imputed

Total

AUC (95% CI)

PPV

1.00E-04

105

233

338

0.803 (0.78-0.82)

0.74

1.00E-05

18

51

69

0.694 (0.67-0.72)

0.62

1.00E-06

6

18

24

0.650 (0.63-0.68)

0.60

 

 

 

 

 

 

 

 

Conclusions                                                                                                           

We have demonstrated that with 336 SNPs we can achieve an AUC of .80, a threshold used for biomarkers. However, for medical diagnosis and treatment an AUC of .95 is desired. We are currently working to identify the p-value threshold and number of SNPs necessary to reach an AUC of .95.

Audience take away:

  • Audience will learn about Genetic Risk Scores and how they are derived.
  • Audience will lean about the positive predictive (PPD) value of a GRS and the reproducibility of these indices in an independent replication sample.
  • Audience will hear discussion about the feasibility and practicality of applying GRS to childhood diseases and this might impact clinical practice and public policy.