I want to analyse my data to see if there is an association between a disease and SNP. How can I perform the analysis to find out this association? I have three genotypes: AA AG GG. G is major allele and reference allele and A is minor allele or disease causing allele. But after performing RFLP and sequencing I got this pattern of genotype in diseased case AA 46/100, AG 53/100 and GG 1/100 While in control I have AA 86/100, AG 14/100 and there is no GG genotype present in control. Due to this I cannot perform chi square test because of small observed value. In this situation how can I perform genotype frequency of both diseased and control and also odd ratio in logistic regression test for association between disease and SNP? Which allele or genotype I used as reference and major allele?
Hello Arifa,
For my answer, I will be assuming you performed a case-control study. The first step in a genetic epidemiology study is the quality control assessment of your data, depending on your methods this would include genotyping call rates, per-sample call rates, gender checks, etc. All these tests can be performed in PLINK, R, or any genetic epidemic software you are most comfortable. Of most importance for case-control studies is assessing the Hardy-Weinberg Equilibrium (HWE), especially in the control group. Any significant deviations would evidence a selection bias or genotyping errors. Therefore the generalizability will be diminished, and your inferences might not be appropriate. Based on your genotype counts, your controls HWE is greater than 0.05 (p ~0.45) so no issues there. Secondly, using any software, you could screen for the types of inheritance patterns. I recommend PLINK or R. Nevertheless, I have attached some basic layouts if you choose to do the bivariate calculations on your own. In general, there are three basic types of analyses you could perform: 1) Genotypic test, 2) Allelic test, 3) Dominant/recessive. There are others but these are the startup points. If the sample size is small, especially if any counts of events in a 2x2 table cell is lower than 5, the computed p-value from a Chi-square distribution is not adequate. A Fisher exact test is the preferred approach. You can calculate ORs and significance values for any of the three modes of inheritance. However, the last step would be adjusting for clinically meaningful variables to the field or those variables in your dataset that are significantly associated with your outcome (esophageal cancer). You could do this with logistic regression modeling. Traditionally, the most frequent allele (major allele, the wild-type allele, etc.) are set as referenced in the regression models. All effects are interpreted based on this allele. The form of your regression would vary depending on the type of inheritance pattern you choose for your data. Again, I have attached some figures of this.
Hopefully, this is of help to you,
Andric
- Badges
- Science method