It migh be helpful to follow an integrative approach, as GWAS would help you get population specific associations, Candidate gene studies would provide significant insights about how pathogenic is your dataset?, so both would enforce your genetic dataset strength as well as relatedness to your population and disease.

For that point, bioinformatics could help you may use DisGenNet tool to build a primary network, but other tools like SNPInfo could help you link genetic associations out of GWAS to functional level signifcance of your genes.

So starting with getting expression data from a database like Genotype-Tissue Expression (GTEx) project, database of Genotypes and Phenotypes . then link the functions of your proposed genetic set to GWAS association through building a network depending on databases like STRINGdb and Programs like CytoScape. That will help you to examine the possible assocaitions and through network analysis you could filter you dataset in a helpful way.

Dear Rahul,

Regarding the fact that you've identified more genes than expected - some records comprise not one risk gene assigned to a disease, but rather a larger region.

Still it is not clear for me which genes are looking for: risk ones or the genes of the proteins which develop Alzheimer? The first one have their impact for a longer time before diagnostics of the disease, while the second are already those which manifest during the disease.

So getting maximal number of genes, actually, will reduce the reliability of the PPI you construct, because for many of the interactions you will not find pairs or they will not be significant as the proteins works in different time.

Probably, you should separate them following candidates genes idea, as well as taking into account are you looking for risk genes or disease related ones.

Good luck!

Dear Dr. Levitchi,

I agree with you. I am looking for risk genes. Do you have any suggestion regarding risk genes figure out ?

I am working with risk genes for the obesity, as risk factor of cardiovascular diseases. I used GWAS by traits associated with the obesity, both anthropometric and biochemical. I generated the table with all related studies. Then I was filtering the data. For me it was important to obtain results obtained or validated on Caucasians. After this I understood that an ontology of traits would be necessary to exclude overlappings, e.g. results for 'lipid' trait and 'HDL cholesterol', excluding the one referred to more general trait in the ontology. This gives you one more outcome - you will know which pathways to analyze, so have hints on involved proteins! Then you may turn your attention to the candidate genes, associated genes etc, in order to check what might be missing. And complement ( but dont forget to to mention the source of each gene in your list, might be helpful in some cases). If you have your experimental data, you may try to "validate" them. Or you can check for microarray or RNAseq data, in order to check your gene list there and identify if reliable data are included regarding gene co-expression.

So, it may make your analysis heavier, but still you need a step to prove what you've obtained.

Good luck