I am following a project about the potential impact of bi-allelic SNPs and clinico-pathological features as well as prognosis of thyroid cancer. Importantly, only affected patients have been included in the study.
These are my questions:
- how to calculate the sample size which is sufficient for achieving an adequate statistical power (I think 80%). I think that it should be calculated for each analyzed SNP starting from the minor allele frequency.
- how to test the association between each SNPs and disease features as well as disease outcome (continous and categorical variables).
- Can you suggest me a freely-available software package specific for makig these calculations ( I read about PLINK and R-packages). Otherwise, May I use some more common softwares as well as SPSS or Graphpad?
Best regards
Hi Vicenzo,
1. To calculate sample size, you can download G power 3.1 from this site http://www.gpower.hhu.de/en.html . It has a user friendly handle to compute your power of the study.
2. For this you need make an input file containing your data. This depends upon the software you want to use. I have used SPSS. For this you need to create a spread sheet of SPSS with headers including the various SNPs and covariates like disease phenotypes/ symptoms, age , gender, BMI etc. Put a numeric value for each category. Such as, for genotype AA=2, AG=1, GG=0 if you follow a suit dominance in case of A allele, similarly, Male=1, Female=2, Status affected=1 and Unaffected =2 and so on.. Once the data sheet is ready, you can choose the test of your interest. For me it was linear regression by keeping Status and SNPs as dependent variables and age and gender as covariates.
3. SPSS is not freely available. PLINK and R softwares are freely available but I find them computationally rigorous.
For calculation of sample size this online calculator http://pngu.mgh.harvard.edu/~purcell/gpc/ is helpful. You could do the calculations for different MAFs and see how many samples you would need.
Regarding software, I think it depends on the number of SNPs that you want to test, PLINK and several R packages (e.g. GenABEL which has a great tutorial, or if you have imputed data you should use ProbABEL) are more convenient if you want to test more than just a few SNPs as this software can do the calculations for all SNPs at once. With these packages you could do logistic regression analysis (for dichotomous variables), linear regression analysis, and the ABEL packages also include functionality for Cox proportinal hazards models in case you want to analyze your SNPs in relation to survival data.
Hi Vicenzo,
1. To calculate sample size, you can download G power 3.1 from this site http://www.gpower.hhu.de/en.html . It has a user friendly handle to compute your power of the study.
2. For this you need make an input file containing your data. This depends upon the software you want to use. I have used SPSS. For this you need to create a spread sheet of SPSS with headers including the various SNPs and covariates like disease phenotypes/ symptoms, age , gender, BMI etc. Put a numeric value for each category. Such as, for genotype AA=2, AG=1, GG=0 if you follow a suit dominance in case of A allele, similarly, Male=1, Female=2, Status affected=1 and Unaffected =2 and so on.. Once the data sheet is ready, you can choose the test of your interest. For me it was linear regression by keeping Status and SNPs as dependent variables and age and gender as covariates.
3. SPSS is not freely available. PLINK and R softwares are freely available but I find them computationally rigorous.
a) sample size calulation: quite some packages are freely available for genetic studies, e.g. http://pngu.mgh.harvard.edu/~purcell/gpc/ , http://www.sph.umich.edu/csg/abecasis/CaTS/ , or http://biostats.usc.edu/Quanto.html . One would have found any of these by google search
b) you have given the answer yourself. Getting accommodated to PLINK is definitely recommended. Besides: for doing single marker studies you can of course use your favorite package
I greatly thank all the colleagues who have kindly given an answer to my question. I have not a MAC-computer, so I can't use PLINK. I have downloaded R-softwares with the packages SNPass and GENabel but it was extremely difficult for me to let this software work. However I will try, and I will let you know....
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics