Ch 12: extract 2

Gigantic lumbering robots, and other fantasies

I realise it is going to take about 10 days to put this chapter up. Having started I think I will carry on.

As Wendell Berry says, ‘we should banish from our speech and writing any use of the word “machine” as an explanation or definition of anything that is not a machine. Our understanding of creatures and our use of them are not improved by calling them machines.’ This section is on the metaphors of the robot and the computer in biology.

From robots to computers – bypassing all that matters on the way

‘Today, the genetic program is everywhere’, writes philosopher of biology Daniel Nicholson:

It appears in biology textbooks, such as when organisms are said to be ‘governed by the laws of physics and chemistry as well as a genetic program’.^[1] It is present in the technical literature, such as when claims are made that ‘a cell can be seen as a computer (a machine expressing a program)’.^[2] And it is also prominently featured in works of popular science. In one of the most memorable phrases of Richard Dawkins’ hugely influential The Selfish Gene, organisms are described as ‘survival machines – robot vehicles blindly programmed to preserve the selfish molecules known as genes’.^[3] Even some philosophers of biology, who should probably know better, have enthusiastically defended the reality of the genetic program.^[4]

And Nicholson adverts to an egregious example:

The genes program the embryo in the same way that the hardware in an automobile assembly plant’s robots’ central processing units realize a program that enable [sic] them to weld chassis without human supervision and more accurately than any human can perform this task. This latter thesis requires no philosophical defence. It is obvious to anyone who examines the available accounts of the embryological development of, say, the Drosophila [fruit fly].^[5]

There are many problems with this (and some of them, as I will show, are illuminated precisely by examining the development of Drosophila). ‘In reality’, writes Nicholson, ‘DNA is only functional when it is embedded in the context of an already present, intricately organized cell’:

It is only in the presence of a pre-existing cellular apparatus that any talk of ‘gene action’ can even make sense. And what is more, the origin of that cellular apparatus cannot be traced back to the genes … many cellular structures such as membrane-bound organelles, as well as the plasma membrane itself, are the only template for their own replication; that is, they are the only source of ‘information’ for their own three-dimensional structure.^[6]

And the distinguished Harvard geneticist Richard Lewontin comments:

Genes are said to be ‘self-replicating’, they engage in ‘gene action’, they ‘make’ proteins, they are ‘turned on’ or ‘turned off’ by ‘regulatory’ DNA. But none of this is true … Genes ‘do’ nothing, they ‘make’ nothing … DNA is among the most inert and nonreactive of organic molecules.^[7]

Indeed, it’s not just what genes don’t ‘do’: it’s what they don’t ‘know’. There is not anything like the necessary set of instructions in them. ‘It is commonly stated that the genome incorporates a Bauplan, an architectural plan or blueprint of the body’, writes immunologist Werner Müller in his classic textbook, Developmental Biology. ‘Actually, this is not the case: the genome is not a sketch or design of the finished body.’^[8]

Many, many other elements than the genotype go to make the phenotype – the organism we can see. As well as DNA, it is obvious that chromatin, RNA, cellular signalling pathways, selective modes of protein translation, and the membrane architecture of the cell all contribute, for a start. The proper response to the news that we share 98–99% of our genes with a chimpanzee, or 50% with a banana, is not that we must be barely distinguishable from chimps and not far from being bananas, but that the determining role of genes as such is more limited than we thought.^[9]

Please note, I am not for one moment denying the importance of heredity: it is very important indeed. I am merely suggesting that it is far more complex than we are led to believe. You cannot consider genes in isolation from many other aspects of the organism. ‘Genetic change’, writes Shapiro, himself a geneticist, ‘is almost always the result of cellular action on the genome.’^[10] Not, then, solely, or even principally, the action of the genome on the cell. But more importantly, heredity is more than genetics: it passes from whole organism to whole organism, not just via DNA – exactly how, we do not know.

It is quite impossible for genes to ‘programme’ the making of an embryo. For a start, there is nowhere near enough information contained in genes. Consider the human brain, never mind the whole human body. With an estimated 100 billion neurones, a quarter of a million of them are created on average during every minute of the nine months of gestation, at times more. Each of these neurones, needless to say, must then make thousands of connexions with other neurones, and all this must end up in a minutely specific, hugely complex architecture with every part of the whole in exactly its required location. Where was that information in the ‘programme’ that DNA is said to enshrine?

We have an estimated 26,000–30,000 genes (what counts as a gene is not a cut and dried matter, but to a certain extent a human convention, and involves elements of judgment). But a blind, millimetre-long roundworm, Cænorhabditis elegans, with only 959 cells in total already has over 19,000. Indeed the pea aphid has 34,600 genes; and Daphnia pulex, a water flea common in lakes and ponds around the world, and about the size of a full stop on this page, has 39,000.^[11] A species of amoeba has the largest recorded genome, 670 billion base pairs, roughly 200 times the size of the human genome.^[12]

And then, only a tiny fraction, about 2%, of our DNA consists of exons, those regions of a gene that actually contain the information required to encode a protein.^[13] With 1,200,000 proteins being produced, it is quite clear that any one gene must code for many different proteins, depending on the context.^[14] The ‘same’ DNA proteins, with the same amino acid sequences, can, in different environments, ‘be viewed as totally different molecules’, with distinct physical and chemical properties.^[15] Protein synthesis is subject to epigenetic influences to such an extent that the same gene ‘blueprint’ can create 2,000 or more variations.^[16] Epigenetic changes are those that are acquired during the lifetime of individuals – from particular experience – and alter gene expression in any one of a number of ways in the next generation. This means that evolution is much smarter than often portrayed. It doesn’t have to wait millions of years, drumming its fingers, waiting for randomness to catch up with experience.^[17]

The other 98% of the genome used to be referred to as junk DNA, because it seemed meaningless, consisting of endlessly repeated sequences that code for nothing. These sequences are no longer considered junk, simply by deduction from the fact that they make a difference – but how they make a difference is not yet known.

Indeed, what is a gene? According to the geneticist Rick Young at MIT, when he first started teaching in the mid-1980s it took him about two hours to teach raw undergraduates what a gene was and how it worked. Today he and his colleagues need three months of lectures to convey the concept of the gene to graduates. Karola Stotz and Paul Griffiths collected 14 atypical but real genetic arrangements and asked biologists to decide whether each represents one, or more than one, gene.^[18] For example, in one case a protein is assembled when four different RNA molecules, made from DNA scattered over 40,000 base pairs, are assembled into one transcript. They found that though the biologists were fairly evenly split, ‘hardly any confess that they don’t know’.^[19] Instead of discrete genes mass-producing identical RNA transcripts, ‘a teeming mass of transcription’ converts many segments of the genome into multiple RNA ribbons of differing lengths. Some of these transcripts come from regions of DNA previously thought not to contain protein-coding genes. The genome contains many overlapping transcripts, genes within genes, and even genes that seem to be controlled by regulatory regions from another chromosome.^[20] ‘Discrete genes are starting to vanish’, according to Roderic Guigó, Director of the European Research Council’s Centre for Genomic Regulation.^[21]

In sum, the extraordinary fidelity and robustness of organismal development is not attributable to a programme, an algorithmic sequence of predetermined steps that must be carried out in a certain order, as with a machine. The cell is constantly monitoring, promoting, inhibiting (including, as necessary, inhibiting its inhibitors), repairing, splicing and rewriting DNA. Each ‘step’, as we are accustomed to think of it, is not computed from the immediately preceding one, but is part of a process in which information comes from many sources, a process that can’t properly be thought of as decomposable into steps at all. The development of an embryo requires an unimaginably intricate series of local interactions, not specifiable by a single global control mechanism. ‘The genetic program does not explain development’, writes Nicholson; ‘it merely black boxes it’:

Embryogenesis is much too complex and much too reliable in its outcome to be specified by a program. If development were a program, the fertilized egg (not to mention the genome) would never succeed in computing its output.^[22]

That a purely deterministic programme would be prone to error might seem counterintuitive; but when one takes into account the degrees of freedom involved in the process when compared with the rigidity of a manmade machine, and the degree of complexity involved in constructing an embryo, it becomes apparent. Dupré explains why any such deterministic programme would fail, using an analogy. He points out that the precision with which fantastically complex processes are carried out, often by different paths on different occasions and under different circumstances, necessitates an overall goal, not a set of deterministic steps – which would be far too likely to go wrong:

If I ask someone to go to the shop and buy me a loaf of bread, and they agree, I am fairly confident that the outcome will be as I intend. If I provide a deterministic programme – take 12 paces north-west, raise hand, turn knob, push, etc, there are too many unanticipated interventions that can derail the process for me to have much hope of success. Teleology is much better than deterministic causation at getting things done, and development is much too reliable to be seen as anything but teleological.^[23]

I will return to teleology in due course. Certainly, the levels of accuracy achieved in the replication of the genome are awe-inspiring and attest to a complex and active process of monitoring of the genome by the cell, resulting in a three stage programme: (i) selection of the correct DNA nucleotide by the replicative DNA polymerase; (ii) removal of any mis-inserted nucleotides by a proofreading activity associated with the polymerase; and (iii) post-replicative correction of polymerase errors that have escaped proofreading by the DNA mismatch repair system.^[24] This reduces errors in the transcription of an entire genome of 3 billion base pairs from a few million to less than 1. This is achieved not by the genome but by the system as a whole operating at several levels. According to the biologist Denis Noble, ‘the order at the molecular (DNA) level is actually imposed by higher level constraints’.^[25] Top-down causation is not supposed to happen in the reductionist model.

Nicholson speaks of the genetic program not explaining development, but merely ‘black boxing’ it. Current biology is littered with such ‘black boxes’. Take the concept of ‘emergence’: it is a perfectly useful term for pointing to what Bergson called creative evolution – the extraordinary coming into being of completely novel qualities as a system grows in complexity. The problem arises when we use it not just to point, but as though offering an explanation. This is like Molière’s doctor grandly explaining that opium makes one sleep because of its ‘dormitive properties’. Speaking gravely of emergence is fine as long as you do not imagine you have explained anything thereby. You have merely sidestepped another problem for reductionism. The rabbit ‘just did’ emerge from the hat. What was to be explained has just been rephrased, and voilà – another mystery solved.

The idea that the development of an individual organism is the unfolding or execution of a genetic program suggests a kind of simultaneous myopia and tunnel vision, lacking both depth and breadth of insight. As Levins and Lewontin point out,

The organism is the consequence of a historical process that goes on from the moment of conception until the moment of death; at every moment gene, environment, chance, and the organism as a whole are all participating … Natural selection is not a consequence of how well the organism solves a set of fixed problems posed by the environment; on the contrary, the environment and the organisms actively co-determine each other.^[26]

The shape that an organism takes emerges from its own physiological states, as much as from genes, and these are as much under the influence of apparently ‘foreign’ organisms as of the organism itself. Referring to the impossibility of distinguishing neatly the intestines from their exceedingly numerous and complex inhabitants, biologist J Scott Turner points out that ‘it is physiology more than genetics that forges the designed gut.’^[27]

Even within the organism, genes are only one part of a bigger picture. The cell context – which so radically changes the way in which DNA will act, and which itself reads, interprets, and deploys, according to need, whatever there is in the DNA to read – is not itself specified by the DNA. ^[28] The assumption that it was would be like Chomsky asserting nonsensically that language springs from the child’s brain without interaction with the environment. Philosopher of science Richard Burian writes:

Without knowing an enormous amount about the contexts in which a sequence and its products are placed – contexts that vary enormously in ways that cannot be predicted from complete knowledge of the genome – there is, in general, no determinate answer to the question, ‘What does this sequence of nucleotides do for this organism?’ Thus, if one tries to proceed strictly from the genome up, one cannot predict, in general, what effects sequence-identified genes will have on the organism and how they will affect its fitness.^[29]

DNA is itself a complex that is twisted in three-dimensions in a way so intricate, and economic in achieving multiple ends simultaneously, that it almost defies belief, so as to promote, bring together, or alternatively shelter from contact, regions of the molecule and their encoding capacity. The structure and its manipulation are at least as informative as the string of DNA itself. The molecule is a three-dimensional entity, not just an abstract two-dimensional string of symbols such as a computer might read, a fact which tends to be overlooked when speaking of ‘code’.

The cell nucleus, which is around six millionths of a metre in diameter, contains two metres of DNA, a feat which is ‘geometrically equivalent to packing 40 km (24 miles) of extremely fine thread into a tennis ball.’^[30] That’s not all, since the 46 separate chromosomes (each averaging, if we continue the analogy, the equivalent of over half a mile long), have to be kept distinct and functional, not hopelessly entangled.

Enzymes known as topoisomerases, whose task is to help resolve complexities in the over-winding or under-winding of the chromosome, demonstrate ‘a spatial insight and dexterity that might amaze those of us who have struggled to sort out tangled masses of thread’, comments Talbott. He continues:

these enzymes manage to make just the right local cuts to the strands in order to relieve strain, allow necessary movement of genes or regions of the chromosome, and prevent a hopeless mass of knots. Some topoisomerases cut just one strand of the double helix, allow it to wind or unwind around the other strand, and then reconnect the severed ends. This alters the supercoiling of the DNA. Other topoisomerases cut both strands, pass a loop of the chromosome through the gap thus created, and then seal the gap again. (Imagine trying this with miles of string crammed into a tennis ball!) I don’t think anyone would claim to have the faintest idea how this is actually managed in a meaningful, overall, contextual sense, although great and fruitful efforts are being made to analyse isolated local forces and ‘mechanisms’.^[31]

Don’t forget that this is not something that DNA itself can programme: it depends on the interaction of a large number of factors. Even then, according to a genetics research group at Holland’s Hubrecht Institute,

not only active, but also inactive, genomic regions can transiently interact over large distances with many loci in the nuclear space. The data strongly suggest that each DNA segment has its own preferred set of interactions. This implies that it is impossible to predict the long-range interaction partners of a given DNA locus without knowing the characteristics of its neighbouring segments and, by extrapolation, the whole chromosome.^[32]

Note, the whole chromosome: and not only that, but the whole cell. And ultimately the whole organism. None of this is reminiscent of how a machine works, nor is any of a whole bunch of factors I will now come on to consider.

In doing so, I have been enormously inspired by, and am greatly indebted to, the illuminating work of philosophers of science John Dupré and Daniel Nicholson – especially, but by no means only, by their edition of a collection of papers by over 20 scientists and philosophers, the most thorough examination of processual philosophy in relation to biology to date, Everything Flows: Towards a Processual Philosophy of Biology, which the editors generously allowed me to read ahead of publication.^[33]

Next up, some of the many reasons an organism is not at all like a machine …

---

^[1] Hartwell, Hood, Goldberg et al 2011 (3).

^[2] Danchin 2009 (3).

^[3] Dawkins 1976 (x).

^[4] Nicholson 2014.

^[5] Rosenberg 2005 (345).

^[6] Nicholson 2014.

^[7] Lewontin 2000 (xii–xiii).

^[8] Müller 1997 (2).

^[9] I am neither endorsing nor dismissing these claims about shared genomes, which are open to dispute.

^[10] Shapiro 2009.

^[11] Pennisi 2009.

^[12] Friz 1968. The precise figures should be treated with caution.

^[13] Shapiro 2013.

^[14] Burian 2005 (237).

^[15] Rothman 2002 (265).

^[16] Bray 2003; Schmucker, Clemens, Shu et al 2000.

^[17] See, eg, Jablonka & Lamb 2015.

^[18] Stotz & Griffiths 2004.

^[19] Pearson 2006.

^[20] Spilianakis, Lalioti, Town et al 2005.

^[21] Pearson 2006.

^[22] Nicholson 2014 (167).

^[23] Dupré 2005 (206).

^[24] Fijalkowska, Schaaper & Jonczyk 2012. See also: Schaaper 1993; and Kunkel 2004.

^[25] Noble 2017.

^[26] Levins & Lewontin 1985 (89).

^[27] Turner 2007 (134).

^[28] Fabris 2018 (246 & 250).

^[29] Burian 2005 (256).

^[30] Alberts, Johnson, Lewis et al 2008 (202).

^[31] Talbott 2010a (13–14).

^[32] Simonis, Klous, Splinter et al 2006.

^[33] Nicholson & Dupré 2018.