The more sophisticated a society, the more highly specialised the professionals. To paraphrase Adam Smith, specialists can maximise value-addition by developing greater expertise. So, a deep, narrow knowledge base is useful, especially in academics.
Unfortunately, extreme specialisation can cause blind spots where inter-disciplinary skills are required. Among the most tragic are the “dropped catches” of medical history. Take puerperal fever. This is potentially fatal septicaemia contracted by women during childbirth. The usual source of infection is a doctor with dirty hands.
For centuries, puerperal fever caused high maternal and infant mortality. In 1844, a statistically aware doctor, Ignaz Semmelweiss, suggested physicians wash hands frequently. That single measure slashed infection rates by 85 percent when implemented. Sadly, it took two decades before Semmelweiss’ suggestion was widely accepted.
Medicine and mathematics used to be totally divorced. Yet, medical research demands large data-set collections and rigorous statistical analysis. If doctors had been mathematically aware, correlations like puerperal fever with hygiene, malaria with stagnant water, etc., may have triggered faster insights as to causes.
Modern medicine, of course, uses statistics, and impressive work is centred on large healthcare studies. Associated areas like genetics and drug research are driven by combinatorial number-crunching. They require juggling molecular and genetic sequences in mind-bogglingly large sets.
Medical research can fruitfully draw from the apparently unrelated discipline of search and market research by adapting the latter’s data-mining algorithms. Retailers like Amazon and Netflix have huge computational models powering recommendations like “Customers who bought this item, also bought (other items)”. These algorithms can deduce, for example, that moderate red-wine consumers don’t buy beta-blockers — ergo, a link between red-wine and healthy hearts.
Many spectacular insights could result from such an interdisciplinary nexus. One example was cited by William Langston, director of The Parkinson Institute and Clinical Center (sic). In October 2009, the New England Journal of Medicine (NEJM) published a study examining Parkinson’s Disease and Glaucher’s Disease (a mutation where fat builds up in internal organs).
The NEJM examined 5,691 Parkinson’s victims across 16 global centres before stating that Parkinson’s sufferers were five times as likely to have Glaucher’s. Langston says a similar conclusion can be drawn in 20 minutes by data-mining the 50,000 DNA profiles registered on 23andme, a biotech company that offers online DNA diagnosis services.
23andme (the name refers to chromosomes) is a leading light in the brand-new industry of off-the-shelf DNA research. Charging fees ranging from $99 to $499, it offers services such as ancestry research, and testing for DNA markers associated with diseases and conditions as varied as diabetes, arthritis, cancer and migraines.
Anyone can offer DNA samples for analysis. 23andme will locate any relatives and ancestors registered on their database, as well as flag possible disease markers in the DNA and offer lifestyle advice to mitigate those genetic risks.
It is at the cutting edge of data-mining in part because Google is an original investor. The co-founder and CEO, Anne Wojcicki, is married to Sergei Brin, who has a great deal of skin in this game because his DNA suggests that he is genetically vulnerable to Parkinson’s.
There are other players in the game as well. As such services get cheaper and become more popular, 23andme and its peers will have exponentially more data to slice and dice in models. That they will find new correlations between health traits and DNA markers is a given. It will then be up to the bio-scientists and medical experts to find the underlying causes.
This turns the traditional scientific method on its head. It seeks to find correlations before it even looks for causation. It may carry its own pitfalls since market basket analysis can often throw up spurious patterns. But it’s also likely to turbo-charge research across several disciplines. The medical market basket researcher will soon be a specialisation, in and of itself.
