GeNORM analysis is the best way to select suitable HKG. Try to avoid normalisers that are involved in the function you are changing with treatment (ie metabolic genes in metabolic cell types)
I agree with James Brown. We did a study once examining the effect of alcohol on macrophages. We used initally GAPDH as the normaliser. Bad choice. GAPDH expression is affected by alcohol. We switched to actin and all was fine.