In October of 1992, genetics researchers made a potentially groundbreaking discovery.
Their findings were published in Nature, and stated that a genetic variant in the angiotensin-converting enzyme ACE appeared to alter an individual’s risk of having a heart attack.
The study involved a total of over 500 individuals from four categories.
After publication of these results, there was initial excitement because this same polymorphism was associated with diabetes and longevity, but after inconclusive replication studies, this excitement swiftly transitioned into disappointment.
In 2000, 8 years after the initial report, a large study involving over 5,000 cases and controls found no detectable effect of the ACE polymorphism on an individual's risk of heart attack.
Meanwhile, the same polymorphism had been detected in dozens of other association studies linking it to a wide range of genetic traits ranging from obstetric cholestasis to meningococcal disease in children.
It’s not rare for initial reports of associations between candidate genes and complex diseases to fail to replicate when further studies are conducted.
Common genetic polymorphisms have an insignificant impact on the risk of disease and sample sizes are often too small to prove anything.
Detecting these subtle effects requires studies involving not just dozens or hundreds of individuals, but thousands or tens-of-thousands.
In small studies, investigators often look at only one – or at most, a few – variants in a single gene. They then subset the data (splitting males and females, for example) to find “significant” results in a specific subgroup.
However, this, combined with a tendency to be biased towards positive results rather than highlighting negative results also, results in a conflicted and confusing body of literature which has ultimately slowed down progress in this area of research.
More recently, the medical genetics community has identified thousands of associations between genetic variants and disease that can be proven consistently and robustly.
This is down to innovations in genome-wide association studies carried out on thousands of individuals.
Not only can these studies detect tiny effects, but they’re not constrained to a particular starting hypothesis regarding specific parts of the genome being associated with a particular disease.
Despite this progress, researchers continue to make this two-decade-old mistake even today. This can be seen in the paper in question by Alex Kogan and colleagues.
This study looked at the candidate gene (the oxytocin receptor) and tested for association between a genetic variant in this gene and a trait called prosociality in a sample of 23 individuals.
If the effect sizes of genetic variants on relatively well-defined traits like diabetes and heart attack are small, the effect sizes of genetic variants on less well-defined traits like prosociality must be even smaller.
Unravelling the genetic basis of variation in more subtle human behavioural traits will be a fascinating process, but it will require adhering to rigorous procedures for study design and statistical analysis, as followed by most large-scale disease genome-wide association studies - but all too often ignored by behavioral geneticists.