Selection of risk associated genetic variants (SNPs)

To deliver only the highest quality results for our customers, deCODE applies strict standards and validation guidelines in the selection of risk markers, or SNPs, used in our tests.

Replication in independent populations

Genetic risk variants are most commonly discovered in case-control studies, in which variations in the genome are measured and compared between a set of patients with a given condition and a set of controls. It is important that the association of SNP markers with a particular disease be widely replicated in independent populations from different medical centers or countries. This is done to ensure that the initial observation is applicable beyond the original study population and to weed out any false-positive risk associations. Since there is a degree of uncertainty in the risk estimate in any study, the definition of a successful replication is one in which any difference in the estimates is within the statistical confidence intervals. We accept only studies that are statistically well powered and published in respected, peer-reviewed scientific journals. When results from multiple populations are combined, the odds-ratio is either derived using a standard statistical procedure based on the Mantel-Haenszel model or the most informative study is taken as a reference.

Replication in at least two independent, equivalently-defined populations of patients and controls is the most important principle of clinical validation.

Continental ancestry

Most of the disease-associated markers currently used to assess risk have been first discovered and replicated in populations of European descent, though a steadily growing number have also been replicated in cohorts of other contintental origins. Since the risk of a given variant can differ substantially between continental ancestries, independent replication and risk assessment must be carried out for each ancestry. Thus, for all diseases, our reports specify which markers have been validated in which major continental ancestry groups,such as Europeans, Asians, Africans and Hispanics.

Identification of the disease associated risk allele

In some publications, the identity of the SNP risk allele can be ambiguous. This problem arises as a result of certain SNP variant combinations (A/T and G/C); of the reverse complimentary nature of DNA; and of the fact that the reference genome sequence may not have been stable when the SNP was defined. If between the time of discovery and the time of application in a test a particular region of the genome has been reversed in the reference builds of the genome, there is a great likelihood that risk alleles within that region have been switched. The so-called TOP/BOT method, defined by Illumina, remedies this problem by using the actual SNP polymorphism (the contextual sequence surrounding the SNP) to designate the allele and strand. Nevertheless, selection of the correct risk allele is a major quality control step in the development of our genetic tests. We compare the reported allele frequencies in publications with information based on our measurements in our extensive proprietary research database and with the public HapMap database. The SNP alleles reported by us are, however, identified according to the dbSNP strand definition of the marker.

Selection of surrogate markers

In some instances, when SNP chips are used for genotyping, the disease associating marker described in the literature is not present on the chip. Often, a surrogate marker that correlates strongly with the risk marker can be used to assess the risk. Only markers that have excellent correlation with the reported disease associated marker (r2 greater than 0.98 based on the appropriate ethnic dataset) are used. In some publications, specific odds-ratios for the suboptimal surrogates are reported as well. In such case, these odds-ratios are used to determine the risk; otherwise, the odds-ratios are assumed to be identical. In selected diseases, deCODE uses a non-chip based technology (e.g. the Nanogen Centaurus assays) to capture key risk SNPs that are not properly represented on the chip platform.

This content was last reviewed on January 25, 2011.