What is Health - Part One
5 Apr 2024
AI won’t fix medicine if we’re using 21st-century technology to ask 19th-century questions
Part One
One of the most influential books I remember reading was Dragons of Eden by Carl Sagan. I read it when I was in my early teens, given to me by the mother of my best friend from childhood days. She handed it to me without great fanfare and I devoured it over the next week. I can trace my fascination for biology and the mind back to that book. The next one that I read was The Blind Watchmaker by Richard Dawkins, another seminal book. From then on I was hooked, fascinated by the intricacies of the living world and by our vast — and yet so limited — understanding of what enables beings to be alive.
The paths of biological discovery, it quickly became clear to me, are often meandering, irregular and full of detours. Take the central molecule of heredity, DNA, of course. It was first isolated by the great Swiss chemist Friedrich Miescher in Basel. He did not fully comprehend what he had isolated. Famously, in one paragraph, he mused if the biological purpose of this molecule might be the storage of phosphate, which could be released when needed by degrading the long and viscous material he had found. To any modern molecular biologist this must sound like heresy — the idea that the cell would degrade, and thereby lose, genetic information in order to meet its need for an inorganic compound like phosphate, is simply ludicrous. Yet, at the time, his guess was as good as anyone else’s, as nobody before him had ever held pure DNA in their hands and was confronted with the need to explain its existence.
The history of biology and medicine is full of such wonderful anecdotes. Some are quirky and innocent, like Albert Szent Györgi’s riposte to the editor of the Biochemical Journal. Szent Györgi had just described the isolation of another biological compound, a sugar from what he could tell, but couldn’t for the life of him find a proper chemical name for it. So, in true fashion of making a virtue out of a dire situation, he named it Ignose in his manuscript, indicating that he in fact did not know what the compound should be called — while still adhering to the standard and accepted nomenclature of naming sugars with the suffix -ose, as in glucose, fructose, etc.
This naming was not acceptable, though, according to the editors of the illustrious journal. His manuscript was returned, with the note to find a more appropriate name for his newly identified compound. Szent Györgi, mischievous as he was, crossed out all the references to Ignose and simply wrote Godnose over it and sent the manuscript back. Much later, this compound was found to be a vital chemical, now better known as ascorbic acid, or Vitamin C.
Other mistakes were less innocent. The causes of the black plague were thought to be bad air, mercury was used to treat skin ailments, and several other grave misuses of chemicals are well documented.
And it is hard to blame physicians of old who failed to understand that bacteria cause the plague, when they had, in fact, never seen a bacterium. Or to think that mercury was harmful when the ill-effects of this chemical appear only decades later. Now, in our age, we conduct double-blinded randomised clinical trials (RCTs), in order to determine if a treatment is working and does not cause any harm. Still, many side effects exist and the full effect of the drugs we make are often not seen, hidden, or ignored.
Zantac, Thalidomide, Glyphosate — all examples of compounds that made it to market while still causing significant harm. Now, of course, the internet of the medical and drug development sphere is full of renewed hope that AI will allow a much improved way to design and test new drugs. First promising reports are appearing, and the field is filled with hope — if not to say hype.
But again — our framing of the problem is determined by our current state of knowledge into what the problem is. This framing was most succinctly formulated, as I briefly touched upon in a previous post, by Robert Koch and his famous four postulates.
The problem with AI in medicine is that we are using 21st century technology to answer a 19th century question.
The problem is that our understanding of health has largely been determined by the medical profession and the pharmaceutical industry, who, in fact, specialise in disease.
It is no wonder then, that similar to Friedrich Miescher in 1871, who did not know that he was holding purified genetic material in his vials, that the current leaders of the pharmaceutical industry do not know what they are holding in their hands when it comes to medical data. They are using an incorrect framing to address the question of health.
In a recent presentation given in Potsdam, Prof Lothar Wieler, chair for Digital Global Public Health of the Hasso Plattner Institute, cited statistics that showed that, while modern medicine has enabled people to live longer, the age of onset of disease has not changed in the past 40 years. In short, all that modern medicine has achieved, is to — statistically — generate more sick people. This is now a great burden for public health systems in terms of financing and resourcing.
This problem won’t be solved by using AI, if all you are doing is trying to solve the same problem as before, just with different tools. More efficient target identification will still only give you a target. As in: a singular target. For chronic and complex diseases, this won’t work.
We have to move away from seeing health as the absence of disease. Our current framing of health does not work. We need to — quantitatively, based on our understanding of living systems (however incomplete that may be) and based on extensive data that we now have — reframe the question. Health is not the absence of disease.
In Part Two of this instalment I will attempt a first quantitative exposition of what health is, in functional, measurable and actionable terms.