Sunday, August 14, 2011

In Vitro/In Vivo

It often happens that, for ethical or political reasons, someone wants us all to stop doing A, and do B instead.   In addition to the straightforward though subjective arguments, we are sometimes told that B is “safer and cheaper and more effective” than A.
This is disingenuous.  Put on your logic cap -- If that were true, you wouldn’t need legislation to force people to do B.

The latest example is in an article in this morning’s Washington Post.   It’s on the front page of the print edition;  yet, with uncanny regularity, precisely the articles I find most interesting  get buried on the Web site; I don’t see it on their homepage anywhere.  Anyhow, here it is:

We there are informed that “chimp research is waning with the emergence of lower-cost, higher-tech alternatives.”  Oh well then, no controversy, right?  No, huge controversy.  Europe has banned the practice, and American facilities are under intense pressure.

I won’t wade into the substance of the debate, which is fundamentally ethical and thus it is hard to argue the other party around.   The only contribution of an armchair logician is to point out cases where conviction or emotion is dressed up as scientific fact.    As to the thorny complexus of scientific facts concerning medical experimentation, I have no idea; but then neither do the “parade of politicians, activists and famous faces, including former New Mexico governor Bill Richardson and chimpanzee champion Jane Goodall” who “mounted an uprising”.  There is, though, one overarching general point to be made.

Simulation of complex systems -- biological systems especially -- is very difficult.  Findings in the area of nonlinear dynamics have shown some sorts of simulation even of inanimate systems to be futile beyond a point.   To figure out what things can, and what cannot, be realistically simulated, and by what means, is itself a major scientific undertaking, by no means resoluble by pointing to some technique or other as being “higher-tech”.

Real-world problems, apart from their ‘local’ complexity, have indeterminate boundaries:  It’s unclear a priori what might be relevant.  Thus, take the recent snatch of Ben Laden.  As you can read in the current New Yorker,
the op was meticulously simulated and practiced over and over again, using a replica of the Abbottabad compound.  Only, the walls of the actual compound were solid, whereas those of the replica were a chain-link fence; and this turned out to have major aerodynamic consequences, so that our high-tech chopper crashed in its own rotor wash.  Oops.

Biological testing is even trickier.  Take the horrendous case of TGN1412. “The data it had seen showed that TGN1412 had appeared to be properly tested on animals, leading it to give the go-ahead for human trials. The regulator had approved the use of the drug at 1/500th of the dose used in animal testing, ‘so it has a high safety margin built in’ “.  No doubt the predicted risk was calculated with an impressive flourish of high-tech decimal points.

This was the result:

Six men are in hospital intensive care – two of them in critical condition – after participating in trials of a new drug intended to treat chronic inflammatory conditions and leukaemia.

The scene was a living medical hell, say witnesses. After being injected with the anti-inflammatory drug TGN1412, patients began tearing their shirts off, screaming that their heads were going to explode. One patient's head swelled to triple its normal size, and patients were soon passing out, vomiting, or screaming in sheer terror.



You can get the impression that, as science marches on, discoveries become routine, predictable -- perhaps programmable.   And it is true that we have discovered much, even in biology.   But much of the progress has been by serendipity -- the polite name for a happy accident -- and by fiddling around in the lab or the field, and keeping your eyes open.  The discovery of penicillin is the best-known instance. Here is a homelier example:

In 1989, a group of Canadian researchers studying a blood pressure drug were astonished to discover that drinking a glass of grapefruit juice dangerously increased the drug's potency.
They were testing the effects of drinking alcohol on a medicine called Plendil. The scientists needed something that would hide the taste of alcohol so that subjects would know only that they were taking the drug and not know whether they were drinking alcohol with it.
"They thought it was a joke. We had trouble getting it published in a major medical journal."


On economic modeling:
Seldom are the premises of such models examined closely.  Their appeal is in the chrome and roar of the engine, not the velocity or destination.
-- Edward O. Wilson, Consilience (1998), p. 203


By contrast, one area in which computer modeling has been spectacularly successful  is space-flight. 
This makes sense.   The deal here is gravity, and its detailed mathematical study goes back to Newton.  Since the spacecraft itself weighs so little, we are dealing with a one-and-a-half-body, sometimes a two-and-a-half-body problem.   Both masses and speeds are too low to require relativistic corrections.  So we are dealing with just a plain old inverse-square law.  Piece of cake.
As a result, NASA has come up with aero-mathematical tours de force -- dynamical chaos, microsteering, sweet spots in the gravity-landscape…

[Update 17 Nov 2011] A highly interesting new article here:


I have been careless, above, using “model” and “simulation” as quasi-interchangeable.   Here is a distinction -- whether a description of actual current practice, or a proposal for tidying our vocabulary, doesn’t matter;  the distinction is a good one:

The difference between a model and a simulation  might be that a model is for the purpose of understanding the phenomenon of interest, whereas a simulation’s purpose  is forecasting.
-- Russell Standish, Theory of Nothing (2006; 2nd edn. 2011), p. 29


1 comment:

  1. But the chimp model doesn't replicate the human model. Drugs developed using the chimp model have failed in humans because our immune response is difference because of the genetically difference between the two species. At least two Hep-C drugs, tested on chimps, had to be pulled off the market because of lethal side effects. There isn't an exact model that replicates the complicated human system. But why use the chimp if it doesn't work? Answer: Keeps people who've been doing it employed....