Genome-wide association study (GWAS) signals replicated in an electronic medical record (EMR) study at somewhat lower ORs. For each disease, the odds of identifying a particular single nucleotide polymorphism (SNP) involved (column 2) in a carefully controlled GWAS study (red squares) was compared to the researchers' ability to identify it in their EMR study (red diamonds). And although the blue diamonds are a bit to the left of the red squares, most of the blues (except perhaps the three at the bottom) are above the 1.0 OR. AF, atrial fibrillation; CD, Crohn's disease; MS, multiple sclerosis; RA, rheumatoid arthritis; T2D, type 2 diabetes. Reprinted from Ritchie et al with permission from Elsevier.

Using phenotype data from online questionnaires, 23andMe found novel associations between certain SNPs and (A) hair curl, (B) asparagus anosmia, (C) photic sneeze reflex, and (D) freckling. In these Manhattan plots, the non-hits are the low-rises and the high-rise peaks represent areas where there were strong, statistically significant hits, appropriately corrected for multiple hypothesis testing. Reprinted from Eriksson et al.

A single nucleotide polymorphism (SNP) with a very high known association with multiple sclerosis was shown to have possible associations with a number of other diseases including malignant neoplasm of the rectum, renal failure, diabetes mellitus, and others. A genome-wide association study (GWAS) of this SNP for pituitary disorders might have been a hit, but the approached used by Denny et al gets the same testable hypothesis with potentially less effort. Reprinted from Denny et al with permission from Oxford University Press.

In this Manhattan plot, environmental variables are grouped into classes and colored by the class. The strength of the variables' association with increased glucose is represented by the colored markers. Those above the line are the hits, and these were actually replicated in the sense that two independent cohorts were evaluated for the association and both showed significance above the line. Reprinted from Pate et al.

Here, drug interactions (x-axis) range from antagonistic (they actual reduce each other's effectiveness) to neutral (zero) to synergistic (the two drugs enhance each other's ability to kill off an infection). One might expect that synergy would be ideal: let's kill off as much as we can as fast as we can. But the treatment efficacy (blue line) flattens out above a certain level of synergy, while the prevention of multi-drug resistance (black line) continues to go down. Optimal results are found somewhere left of the shaded area. Reprinted from Torella et al.