Genes which are not associated with Alzheimer's

Question

[Paul]

I require a set of genes which are not associated with Alzheimer's.

I got a list of all genes from OMIM genemap2. However, I am unable to put a filter of genes which are not associated with Alzheimer's. Please suggest me a way to find out the genes which are not associated with Alzheimer's.

[ATpoint]

Does the gene need to be expressed in Alzheimer's?

[Paul]

No, need not to be expressed in Alzheimer's @ATpoint

Answer 1

[Michael]

You are possibly doing some sort of machine learning approach and are in need of a negative set. While this is a quite common concern when doing this kind of analysis, you have to understand two important things:

• There is no statistically or logically clean way to ensure that a gene G is NOT associated with a phenotype P, meaning that P occurs independently of G, always. There could be several reasons for a hidden association: G ~ P, e.g. it has not been observed yet because the power of the experiments so far was too low, we have not included the right cohorts (or the sample was not representative), etc.
• Genes (at least if we understand them as database entries, not as more general heritable entities) do not cause disease, most people have all the disease-associated genes in their genome. Certain variations and their alleles will give rise to G ~ P, as it is also visible in OMIM. You need to consider if the feature space of your method needs to cover the variations, too.

If you need to construct a negative set anyway, you might as well select one or many random sub-sets of all Ensembl-genes that are not listed as associated in any database, as any other criteria will be shaky. The set size should be properly balanced for most ML methods.