Linking Genetics to Disease – Mid-Course Adjustments
In their 2012 PLOS Genetics paper, “Comparison of Family History and SNPs for Predicting Risk in Complex Disease”, authors Do, Hinds and Francke predicted the lifetime morbidity risk and heritability of liability risk for 24 diseases.
Number 1 on the list was coronary artery disease with a 40% lifetime morbidity risk and a 49% heritability of risk liability. Last on the list was Crohn’s disease with a 0.5% lifetime morbidity risk and a 56% heritability risk liability.
The implications were obvious. By applying sophisticated statistical methodologies to genomics, metabolomics and other ‘omics fields researchers could rapidly find associations and, perhaps, causation elements for a wide range of disease states.
And, given demand for novel studies with p-values under 0.05, these papers were comparatively easy to get published.
Yes, the quality of those associations is highly dependent on effect size, background noise and sample size.
But hey, published! How do you spell “Tenure?” G-E-N-O-M-I-C-S.
Tangled up in Causation vs Correlation
To feed the explosion of interest in genetic variants and their potential disease associations, biobanks sprang up around the world (in our research we found about a dozen) to collect, store, and use genetic data from thousands of human samples related to diseases (the Finnish National Institute of Health and Care’s biobank collects specimens and measurements from patients with different diseases, the UK biobank collects samples and physiologic measures for approximately 500,000 people).
And the global medical science research community began to move into the brave new world of correlative, ‘omics scientific inquiry. Before 2010, said Nature Magazine citing data from Scopus and the Web of Science, fewer than 100 papers published per year on the subject of genetic associations with disease.
By 2015, the number of such papers had increased to about 200. In 2019, so far, more than 500 papers have been published.
New Nature Article Alleges Gene Based Hack and Sloppy Genomics Research
A December 10, 2019 article in the journal Nature titled: “The Gene Based Hack that is Revolutionizing Epidemiology” describes the downside of this genomic research land rush.
Author David Adam wrote: “As genetic data have piled up, a flurry of Mendelian randomization studies have emerged that don’t make the grade. Some have relied on misleading data, and others have failed to sufficiently test the assumptions on which Mendelian randomization relies. It’s time, many in the field say, to tighten things up.”
Adam put a portion of the blame for this sloppiness on overreliance on what he called a “hack”—Mendelian randomization and, specifically, the increasing habit of researchers to throw crazy ideas against the wall to see there is are ANY correlations between those ideas and genetic variants.
Instead, said Denize Atan, an ophthalmologist at the University of Bristol, genomics researchers should “Have a robust hypothesis and some supporting evidence before you start. You think, ‘Where did they get that idea from?’ It just seems to come out of the blue.”
Added Sonja Swanson, an epidemiologist at the Erasmus University Medical Center in Rotterdam, “It doesn’t take much to just hit the buttons and say, ‘here’s a numeric answer to my question.’”
One study which claimed to have found a genetic link between smoking while pregnant and underweight babies with orofacial clefts, was panned by epidemiologists because the studies were biased towards finding some effects in the genes being examined. The genetic variants the authors used in their Mendelian randomization had not shown up in larger, more comprehensive genomic wide association studies (GWAS).
In other words, the researchers looked at the data narrowly and, in effect, hacked their way to a published paper.
Correlations vs Causation
Each person has approximately 3 billion pairs of personal DNA information. Tools like high-throughput genotyping and precision reference mapping will someday make this data and all of its disease associations available to orthopedic physicians.
It will be critical, therefore, for physicians to keep their BS antennae tuned.
This land rush is happening in academia currently. It is only a matter of time before this information seeps into your clinic and office.
It may get there without FDA approval, clearance or license because it starts with correlative associations rather than causative data.
We’ll just have to be as aware of Mendelian hacks as we are of P-Value hacks.