Is Imputation Compulsory For Association Study?

Question

[kumar.vinod81]

is imputation compulsory for association studies?

My results are completely different in before and after imputation. Which data I should believe for my further work. Should I believe, the imputed data or the data without imputation are more stronger? What is more trustful? I am working in plant so imputation was performed without any reference panel using fastPHASE.
Thanks

[Giovanni M Dall'Olio]

How many missing genotypes do you have? How many markers does the imputation add?

[kumar.vinod81]

Out of total 1,150,000, my data has near 100,000 missing SNP sites. What should I do

Answer 1

[Larry_Parnell]

Of course, the most trustful genotype data are those measured directly. If you have such different association results with genotyped and imputed data, it sounds to me that there was a problem with the imputation or the assignment of imputed genotypes to each subject. Imputation of genotypes can help to fine map the association, but with your results I would consider another mechanism to perform that fine mapping.

[kumar.vinod81]

Thanks, but what are the other mechanism used in such situation ?

[Larry_Parnell]

Plants are easy to make as RILs and so crossing strains may be beneficial. One could sequence across regions to which the trait was mapped. Or use synteny with a another plant - I did this with Arabidopsis to find targets in soybean.