Lifting the curse of the clinical cut-off

09 July 2012

Enhancing the power to identify genes causing phenotypic variation

Twin and family studies have shown that many traits are heritable. The logical next question is of course: which genes are involved? Genome-wide association studies (GWAS) are currently used to find the genes, or genetic variants, that are associated with traits of interest like depression, schizophrenia and Crohn’s disease. GWAS is an exploratory technique: genetic variants across the entire genome are tested for their relation with the trait of interest. Although GWAS, in general, facilitated important discovery of new biological knowledge about diseases, identified variants for psychiatric disorders explain little variation, and insight into the role of genes in highly heritable psychiatric traits remains poor. Low statistical power is seen as the main reason for the failure to locate more variants, and has resulted in a call for larger samples. In their Letter to the Editor, van der Sluis et al show that better use of (available) phenotypic information can also considerably increase the power of GWAS to detect the genes of interest.

Generally, diagnostic instruments are used to obtain a measure of psychiatric disorders like depression and Attention Deficit Hyper-activity Disorder (ADHD). These diagnostic instruments were developed to aid clinicians in deciding whether a person does, or does not, suffer from a certain psychiatric disorder. To this end, diagnostic instruments include items that refer to extreme behavior like “I felt worthless‚, “I had crying spells‚, or “Steals from other children‚. If the number of items that a person endorses (i.e., symptoms that a person displays) exceeds a certain prior established threshold (e.g., 7 out of 15 symptoms), then the person is said to suffer from the disorder of interest. If, however, the number of items the person endorses does not exceed the threshold, then it is assumed that the person does not suffer from the disorder. This way, the score on the diagnostic instrument is dichotomized into what is called a case-control status variable. In GWAS studies, researchers compare genotype frequencies between cases and controls and those genetic variants that show large frequencies differences are assumed to be related to the disorder.

Van der Sluis and colleagues show that this practice of dichotomizing test scores before analysis decrease the power to detect genetic variants substantially. In addition, they show that the power of GWAS can be further enhanced if researchers would use instruments that do not only distinguish well between cases and controls. For example, two persons who both do not suffer from clinical depression can still differ substantially in how often they feel blue or lonely, how often they worry, and how often they feel afraid. This phenotypic variation within the control group is very poorly captured by diagnostic instruments, which were not developed with that
aim in mind. The authors show that adjustment of renowned instruments or development of new instruments that reflect phenotypic variation across the entire trait range (i.e., from not-at-all depressed to very depressed), could increase the power of GWAS, and thus increase the success of gene-finding studies.

Link to paper online:

Supplemental material: vdSluis_MolPsych.pdf (134.5 KB)