Hi Faizal, 

Are you sure you want to do only PCA on this sort of data? I mean, PCA is the simplest of methods and for the kind of data you mentioned, you would be better off with more clustering methods. For starters, PCA has limitations especially with respect to the scalability of variables and also makes debatable assumptions with respect to how the variables are possibly correlated with each other to form a set of values of linearly uncorrelated variables (principal components). Moreoever, the data could be visualized only on a two-dimensional way - a scatter plot. A bubble graph would be ideal in this case.

The most appropriate method for your problem may be multiple imputation (often referred to as K nearest neighbor). This would allow for the flexibility of a distance term. For starters, you would be better off considering the codominants. Here's some useful quick reading 

https://onlinecourses.science.psu.edu/stat505/node/147

https://www.mailman.columbia.edu/research/population-health-methods/cluster-analysis-using-k-means

Here's a couple of papers discussing how they apply K-means clustering on data similar to yours.

http://ieeexplore.ieee.org/document/6755142/

https://www.ncbi.nlm.nih.gov/pubmed/27311823

Now, considering your data has some typical YES or NO kind of binary values, K-means will not be a good technique to choose. You would be better off performing a Hierarchical Clustering which is more memory intensive compared to a k-means clustering. 

And another consideration if some data points fall into more than one variable categories, you might as well consider a fuzzy c-means clustering!!

Good luck with your work. 

Dear Sir

Thanks for your kind reply and guidance. I see the relevant links and that are very informative and helpful but in my case sample size the very small (n=90).

After PCA I correlate the both data and find out the p-value now I am performing the association with the genotyping data and clinical variales. After that, I have the plan to perform the dominant and codominant analysis with these variables.

Will you suggest I am going on the right path?

Thanks