Polite Row About Models in Biology
Total Page:16
File Type:pdf, Size:1020Kb
NEWS AND VIEWS Polite row about models in biology A public debate last week between Stuart Kauffman and John Maynard Smith entertained a London audience but left the status of modelling in biology iII-defined. DISAPPOINTINGLY little blood was spilled Kauffman last week described a simple asked that "biologists should know a limit last week at the Linnean Society of Lon model of this state of affairs - a peptide cycle when they see one" and know the don, where Professor John Maynard with four amino acids for which the possi causes of bifurcation in the evolution of Smith (Sussex) and Dr Stuart Kauffman ble variations are the replacement of one chaotic systems. By way of illustration, he (Sante Fe Institute) had been tempted by amino acid by another. For what it is noted the apparent doubling of the num the London Evolution Group to a public worth, the model is derived from one ber of segments in Drosophila (from seven debate on Kauffman's mathematical mod introduced by Maynard Smith some time to 7 x 2 = 14), suspecting that "Drosophi els of evolution. In the event, they were ago (Nature 225, 563; 1970). To make the la must have been reading my book". restrained by their innate good manners, problem manageable, it is supposed that Maynard Smith accepts the need for a even their mutual liking for each other; the only allowable amino acids in the relatively smooth fitness landscape, as Kauffman was one of Maynard Smith's four-peptide are alanine and glycine. defined, if evolution is to happen. But he first graduate students 30 years ago (and In principle, it should eventually be could not follow Kauffman's supposition was probably as precocious then as he is possible to compile a table of how all pos that self-organizing systems need external still), and even claimed (physician as he is) sible variations affect the effectiveness of information to assemble, and suspected to have cured Maynard Smith of pneumo the peptide in its biological role, and thus that Kauffman's view that evolving organ nia contracted on a visit to Chicago. to map their fitness. Using arbitrarily cho isms influence their fitness landscape with Last week's disappointment is real, sen numbers, computer modelling shows a view to future evolution smacks of the because the advertised debate concerns a the convergence of different peptides to a Gaia hypothesis. The essence of his polite real issue. Kauffman's book The Origin of common end-result. public dissent is simply put: the modelling Order, in which his argument was More generally, it is even now possible of The Origin of Order notwithstanding, rehearsed, caused something of a stir on to demonstrate that there are indeed local "We still do not know whether, if the envi its publication 18 months ago. A "mixed maxima on the fitness landscape, and that ronment were to stay constant, evolution reception" is what the publishers (Oxford the landscape as a whole is the more would come to a stop". University Press) would have said (see rugged as the fitness comes to depend the Meanwhile, it is clear that Maynard Nature 365, 704; 1993). But last week, more extensively on variations at different Smith is not a devotee of Per Bak and his Kauffman would have known in advance sites in the four-peptide. The greater the sandpiles. Among other things, he won that there were some friends among the interconnectedness, the more peaks dered what could be the correlate in the audience. One distinguished British physi appear in the landscape and the more dif real world of the stream of sand-grains cist volunteered that he had found the ficult it is to reach local maxima, let along that eventually make the pile unstable. He book to contain the only comprehensible the global maximum, by random walking also complained of an article in Nature (by account of the mathematics of evolution. on the landscape. this writer) "with punctuated equilibrium Kauffman's argument is straightfor So, says Kauffman, a smooth rather in the title" (Nature 371, 197; 1994). That ward enough. The principle of order is the than a rough fitness landscape is a prereq concerns numerical simulations of evolu self-organizing principle, right? And living uisite for orderly evolution. How can that tionary models in which the fitness of co things are self-organized entities; look be achieved? Co-evolution is one way. ("If evolving species is determined by how much like a cell a synthetic lysosome a frog evolves a sticky tongue to catch random-number generators under partic seems! The principles of neo-Darwinian prey more effectively, the flies it catches ular rules. One conclusion is that it is pos evolution are heritable variation, the dif had better evolve sticky feet.") But evolv sible to write the rules so that the initial ferential fitness of the inheriting individu ing organisms in general must change the state of the assemblage of species does als and then natural selection. So the ruggedness of the fitness landscape their not affect the outcome; another is that central problem, to Kauffman, is to successor organisms inherit. there can be circumstances when the sys understand how self-organization and nat There remain the grand questions - tem becomes unstable and several inde ural selection can operate simultaneously. the great Cambrian radiation of species, pendent species emerge. Of this sim Nobody last week disagreed about the for example; was there something special ulation and the sandpiles (as a model for use of the likely number of viable off about the fitness landscape of organisms real biology), Maynard Smith says: "I just spring as a measure of fitness, but Kauff then? Kauffman is a confessed devotee of find the whole enterprise contemptible". man sees the representation of the the arguments (largely due to Per Bak) of What last week's audience made of the problem as what the mathematicians call a the critical behaviour of self-organized exchange is anybody's guess. Many beers complex problem. There is a "fitness systems, as in the formation of avalanches later, the two protagonists did agree to try space" or "landscape" with as many on the sides of a sandpile to which grains to define what their disagreement really dimensions as there are potentially herita are added one at a time. The sandpile of is. After careful thought, Maynard Smith ble variations in which, as with complex self-organized living creatures must have announced, "You see, Stu, I don't find it problems in physics, there will be a very reached a critical condition in the early interesting". And that, of course, is what large number of local maxima, corre Cambrian, spawning avalanches on its the argument should have been about. sponding to phenotypically advantageous sides, each a new species or a genus. What is the place of modelling in biology? inherited configurations of variations. So Maynard Smith, an electrical engineer Can a model be heuristically valuable how does natural selection, working on by background, a graduate student of even when it entails only a sketchy corre the variations of fitness a group of organ J. B. S. Haldane and a modeller on his spondence with the real world? Maybe isms has inherited, manage to beat a path own account, is not one to deny mathe the Linnean Society should return to that to the global maximum of fitness? matics a place in biology. Last week, he charge. John Maddox NATURE . VOL 373 . 16 FEBRUARY 1995 555 .