Sometimes, scientific discovery is just a matter of sitting down and using the tools of "connected science" already available to us. Stories like this one underscore the need for generalists:

Don Swanson seems an unlikely person to make medical discoveries. A retired but still active information scientist at the University of Chicago, Swanson has no medical training, does no medical experiments, and has never had a laboratory. Despite this, he’s made several significant medical discoveries. One of the earliest was in 1988, when he investigated migraine headaches, and discovered evidence suggesting that migraines are caused by magnesium deficiency. At the time the idea was a surprise to other scientists studying migraines, but Swanson’s idea was subsequently tested and confirmed in multiple therapeutic trials by traditional medical groups.

How is it that someone without any medical training could make such a discovery? Although Swanson had none of the conventional credentials of medical research, what he did have was a clever idea. Swanson believed that scientific knowledge had grown so vast that important connections between subjects were going unnoticed, not because they were especially subtle or hard to grasp, but because no one had a broad enough understanding of science to notice those connections: in a big enough haystack, even a 50-foot needle may be hard to find. Swanson hoped to uncover such hidden connections using a medical search engine called Medline, which makes it possible to search millions of scientific papers in medicine—you can think of Medline as a high-level map of human medical knowledge. He began his work by using Medline to search the scientific literature for connections between migraines and other conditions. Here are two examples of connections he found: (1) migraines are associated with epilepsy; and (2) migraines are associated with blood clots forming more easily than usual. Of course, migraines have been the subject of much research, and so those are just two of a much longer list of connections that he found. But Swanson didn’t stop with that list. Instead, he took each of the associated conditions and then used Medline to find further connections to that condition. He learned that, for example, (1) magnesium deficiency increases susceptibility to epilepsy; and (2) magnesium deficiency makes blood clot more easily. Now, when he began his work Swanson had no idea he’d end up connecting migraines to magnesium deficiency. But once he’d found a few papers suggesting such two-stage connections between magnesium deficiency and migraines, he narrowed his search to concentrate on magnesium deficiency, eventually finding eleven such two-stage connections to migraines. Although this wasn’t the traditional sort of evidence favored by medical scientists, it nonetheless made a compelling case that migraines are connected to magnesium deficiency. Before Swanson’s work a few papers had tentatively (and mostly in passing) suggested that magnesium deficiency might be connected to migraines. But the earlier work wasn’t compelling, and was ignored by most scientists. By contrast, Swanson’s evidence was highly suggestive, and it was soon followed by therapeutic trials that confirmed the migraine-magnesium connection.

From Reinventing Discovery by Michael Nielsen (a past Singularity Summit speaker).

New to LessWrong?

New Comment
8 comments, sorted by Click to highlight new comments since: Today at 11:36 AM

I read about Swanson a few years back and found him interesting; I couldn't find his papers anywhere but a local medical library which wouldn't let me in, so I never found out very much - but I don't remember finding that anyone had replicated his feats. If there is really that much being missed, why aren't there other people mining the literature and striking gold and becoming metaphorically rich?

I'd guess that if Don Swanson's methods are as promising as this book suggests they might are, then eventually, a lot of people will eventually, gradually, adopt them.

Of course, it would happen much faster if people could become literally rather than metaphorically rich by using his methods.

Also, I suspect that the academic rewards system -- mostly set up to provide rewards like prestige and repute within a very narrow speciality -- would have a hard time adjusting itself to accommodate any multidisciplinary, "connected science" approach, even if its benefits and successes are obvious. Is the medical science establishment even capable of rewarding someone like Don Swanson? Maybe that's not fair, because Don Swanson is retired. But what specific academic department in what specific school is going to give some young researcher tenure because of some innovative discovery involving "connected science"? And, in the meantime, why should a young graduate student spend precious time on "connected science" when the academic/research job market is already competitive?

If this approach is as valid as it seems, the powers that be will come around eventually -- but not quickly.

P.S. Certainly, the pharmaceutical companies should be working these methods to the hilt if they're not already (probably they are.) But still, are these comparative literature methods likely to result in a new, patentable drug? The result described above involved magnesium deficiency, which is good to know, but provides no obvious great profits to anybody involved. I'd bet that, in the medical field, at least, most "connected science" results would show enormous health benefits from unpatentable results like eating a balanced diet, exercising regularly, not smoking, and having good genes.

[-][anonymous]12y20

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1459187/?tool=pmcentrez

The above provides a nice summary of related text mining work. It doesn't sound quite as impressive without the tech journalism rhetoric, but the results are comparable.

Data mining in general, or specifically this sort of data mining? I don't know about this specific kind, but the general notion is alive and well.

Also, it could be that doing this sort of thing well is very time-consuming and low-yield - How long did it take him to find that connection, and how many connections could he have found in that same data, if they were there?

Because people are dumb?

More seriously, I suspect a lot of people wouldn't expect themselves to have anything to contribute to fields outside their area of expertise just as they suspect outsiders wouldn't have the ability to make breakthroughs in the area they are "experts" in. This seems like the kind of activity that yields no results at all until you find something important, so maybe it seems like a waste of time?

A vaguely related essay I found through HackerNews: http://psychsciencenotes.blogspot.com/2011/11/theory-and-why-its-time-psychology-got.html . It is also a good piece about why theories are important in science in general.

This feels like it should be a separate post to me.

Here's Michael Nielsen TED talk on open science and the need for scientists to 'embrace new tools for collaboration'

http://bit.ly/usz8pr