ABSTRACT
The identification of environmental and genetic factors that contribute to disease risk requires appropriate statistical methods and software that can integrate different sources of risk, provide statistical assessment of combined risk factors, and facilitate interpretation of this risk. We have developed an R package, REGENT, to calculate risks conferred by genetic factors and multilevel environmental factors. This is performed at a population level, with the option to also analyse individual-level data. REGENT incorporates variability in risk factors to calculate confidence intervals for risk estimates and to classify the population into different categories of risk based on significant differences from the baseline average member of the population. REGENT is an R package available from CRAN: http://cran.r-project.org/web/packages/REGENT. It will be of value to genetic researchers exploring the utility of the variants detected for their disorder, and to clinical researchers interested in genetic risk studies.
Subject(s)
Algorithms , Crohn Disease/genetics , Polymorphism, Single Nucleotide , Risk Assessment , Software , Autophagy-Related Proteins , Carrier Proteins/genetics , Confidence Intervals , Environment , Gene-Environment Interaction , Humans , Nod2 Signaling Adaptor Protein/genetics , Receptors, Interleukin/geneticsABSTRACT
PURPOSE: To develop a method of genetic risk categorization based on the risk conferred by genetic variants and the precision with which risks are known. METHODS: We develop a method for risk assignment based on an "average" member of the population and their genotype, deriving empirical confidence intervals encompassing all relevant sources of variation in disease risk. An individual with risk confidence interval that does not overlap with that of the "average" individual is categorized as having higher or lower disease risk. The method is applied to data sets in Crohn's disease and type 2 diabetes. RESULTS: The proportion of the population assigned to the average risk category depends on genotype relative risk, allele frequency and sample size of the study used to estimate these parameters. For low genotype relative risks or minor allele frequency, little resolution into different risk categories may be possible. CONCLUSION: The utility of a genetic risk variant for risk categorization depends on both the magnitude of the genotype relative risk and the accuracy with which this, and other elements of risk calculation, are known. Genetic risk calculations should include an assessment of the accuracy of the risk estimation.
