The conserved oligomeric Golgi (COG) complex consists of eight subunits organized in two lobes: lobe A (COG1-4) and lobe B (COG5-8). The different functional roles of COG lobe A and lobe B might result in distinct clinical phenotypes in patients with COG-CDG (congenital disorders of glycosylation). This hypothesis is supported by three observations. First, knock-down of COG lobe A components affects Golgi morphology more severely than knock-down of COG lobe B components. Second, nearly all of the 27 patients with lobe B COG-CDG had bi-allelic truncating mutations, as compared with only one of the six patients with lobe A COG-CDG. This represents a frequency gap which suggests that bi-allelic truncating mutations in COG lobe A genes might be non-viable. Third, in support, large-scale exome data of healthy adults (Exome Aggregation Consortium (ExAC)) underline that COG lobe A genes are less tolerant to genetic variation than COG lobe B genes. Thus, comparable molecular defects are more detrimental in lobe A COG-CDG than in lobe B COG-CDG. In a larger perspective, clinical phenotypic severity corresponded nicely with tolerance to genetic variation. Therefore, genomic epidemiology can potentially be used as a photographic negative for mutational severity.