I have problem concluding this data of polymorphism association with a disease. The SNP shows no significant association with respect to (wrt) allelic frequency (P >0.05) but shows a marginal (P = 0.03) association wrt genotypic frequency. How should I draw a conclusion from this data?
Hi
it is a good question. You should keep in the mind that always number of genotypes is more than number of alleles, for example with a rule of thumb for n allele we have n(n+1)/2 genotypes. Therefore in your case one reason can be the further number of genotypes in compare to number of alleles. Also, in association study genotype frequencies are more important than allelic frequencies, because in this study we investigate the association between phenotype of individuals that each one is a genotypes.
I assume that the question is in the context of complex-trait association. Depending on the model by which the genotypes were analysed, the result can be perfectly valid . Allelic association assumes a multiplicative model, whereby each allele increases risk by the same factor. While it is probably the closest to the truth in most complex traits, it disregards threshold effects that would result in a recessive model (one allele has little or no effect, two are required to go over the threshold) and maxing out of the response curve resulting in a dominant effect (one allele already maxes out and the second has no additional effect).
If the genotypic association was done by an agnostic 2x3 design (affected/unaffected vs. 0/1/2 risk alleles) at a diallelic locus, be careful. If the statistical significance is on the heterozygotes being different from the other two groups combined, you are probably dealing with a spurious result. This can happen only in very rare special situations, such as HLA in autoimmune diseases, where the effect of the alleles at a multiallelic locus is qualitative rather than quantitative. For example, DR3/DR4 heterozygosity predisposes to type 1 diabetes more strongly than homozygosity for either allele. This is because different alleles are better at handling different antigens. However, In most cases in the genetics of complex traits, the effect is purely quantitative.
Thank you for your replies Ayoub and Constantin.
I am dealing with a diallelic polymorphism which when assessed with the perspective of allelic frequency doesn't show a significant association with the disease. But the genotypic frequencies were distinct among patients and control groups. Therefore I checked for the different models of association, dominant, recessive, additive. I found that the polymorphism was dominantly inherited and that the risk allele homozygosity incurred higher risk of acquiring the disease. Hence the alleles as per are not associated but maybe a particular genotype can incur sufficient risk in acquiring the disease.
Dear Gargi Nanda,
Be aware that the multiplicative model (allelic) should only be applied if the genotype frequencies of the marker in question adhere to HWE in cases (controls, ideally, should always adhere to HWE). If this assumption is violated, the allelic model should not be applied. For more details, see (Sasieni, P. D. (1997), ‘From genotypes to genes: doubling the sample size’, Biometrics, Vol. 53, pp. 1253–1261).
Thiago makes a good point. Since it is the allelic test that is not significant, you will strengthen your case if you show departure from HWE in the cases.
But my study population (controls and patients) have their genotypes in HW equilibrium.
Not the entire study population, you should only look at cases. Even if cases alone are in HWE, you could still have an effect that significantly diverges from the muliplicative model. If they are not, makes your case stronger.
- Badges
- Science topic
- Similar topics
- Correction