Texture Garden

Tree Shaped Chromosomes

This document discusses the virtues of employing tree-shaped chromosomes in artificial life programs, and programs that use genetic models which include genetic recombination.

This essay was written in 1996. My views on some of these issues have changed since then.


Modern Life is Rubbish
Coloured Bodies

Modern life is rubbish

In the process of presenting evidence for the theory of evolution, some scientists have drawn attention to the way in which some organisms seem to be adapted to their environment in ways that seem to suggest that an historical development has been instrumental in moulding their forms.

Such signs of history have been referred to as Scars of Evolution by Elaine Morgan. The word scars is well chosen as it implies that the signs are almost always maladaptive to some degree or other, and that any rational designer would avoid such features.

Many organisms display signs that one structure has been adapted from an original use to a new role, signs of lock in of arbitrary or non-optimal designs, vestiges of historical structures of no apparent utility, and sometimes complete lack of features which it seems they would find advantageous, but are not suggested by small variations in the embryological development of their forms.

As well as being applied to specific phenotypic aspects of individual organisms, the same technique of examining apparently imperfect adaptations may be applied to the study of the genetic machinery in an attempt to trace its history.

When examining multicellular organisms such as ourselves, one of the curious features is that we would appear to be colonies of millions of bacteria-like organisms. These creatures are all clones of one another, descended from a single cell, and many of them are still capable of asexual reproduction. Amazingly, each of them contains a copy of the entire genetic heritage of the organism.

There are undoubted advantages in storing the information locally to the cell. In particular, translation and transcription into amino acids can occur adjacent to the information store, and when the cell decides to reproduce in mitosis it has all the information it needs to pass to its new sister cell.

It seems that there are also huge disadvantages to using local storage. The sheer quantity of redundant information involved is colossal, when it is considered that there must be a cost in terms of the chemicals, time and energy required to support all the information.

As organisms age their genetic homogeneity decreases due to mutations in the individual cells. In older people this results in cancers, as well as other signs of cellular lack of harmony symptomatic of ageing. Those cells in the germ line responsible for gamete production are affected in the same manner as other cells. With all the copies of the genome available, you might have thought that this resource could be used as a multiple back up device, or as a gigantic error correction system, but in fact cells do not communicate with their neighbours about genetic matters directly at all.

The question arises: would the organism be better off if it could maintain a central genetic database and then network its cells together?

If this proved tp be possible then in principle a whole series of changes could be made. If, for example, the information store could be decoupled from the two important processes of transcription and replication, then a more energy-efficient and secure storage system could be developed.

If cells could communicate genetic information to one another then improved error checking could be implemented and compression techniques would be available if required. Viruses could in principle be prevented from subverting such communication channels by employing encryption techniques, though selection pressures in a natural environment would be unlikely to be sufficient to achieve this goal. Better cell to cell communications could in principle eliminate the physical substance transfer involved in sex.

The kind of communication required need not be fast; indeed the speed of the nervous system would be more than sufficient: genes do not control their hosts in real time but via the slow-motion remote control of protein manufacture. Communications could in principle take the form of local transactions between adjacent cells.

There is no proposal here to defend here the viability or otherwise of such possible alternatives to the current arrangements, but that such possibilities exist and may be greatly superior to existing systems is worthy of consideration.

Because it is at the bottom so to speak of a series of developmental stages, the genetic substrate may be helpfully seen as the most ancient and primitive structure of modern life, and though there is a sense in which it is immensely high technology, there are many ways in which its 'design' seems to bear witness to the nature of life's last common ancestor, rather than modern utility.

It is possible that the existing arrangement exists as a kind of local optimum of a system with much better global maxima. It may be that developmental constraints within organisms have limited them to an existing system which displays all too clearly its historical legacy.

In many systems that have been evolved rather than designed early aspects of the design become in a sense locked in by all the subsequently constructed structures which are built on top of them.

The phrase The bathwater cannot be thrown out because of the baby would seem appropriate in this context. Unfortunately, as time passes the dirt in the bathwater is liable to accumulate as more and more lock ins occur on different levels, until it is difficult to make out the baby at all.

Any watching designers would wring their hands in frustration, desperate to be able to start on a fresh new drawing board, and use the old evolved design as the basis for a completely new organism with all such constraints designed out.

Coloured Bodies

The invention of sex may be a usefully considered to be a major landmark in the development of modern organisms.

One of the reasons for this is that it may serve to mark the end of the evolution of the main structure of the genetic code. Any major structural changes in the genome are likely to render the organism which hosts them to be incompatible with other members of its own species when it comes to mate with them, rendering it sterile. Naturally, any sexual organism incapable of interbreeding with its fellows is genetically doomed.

If a new variation upon a genetic mechanism is good enough that its organisms can afford to abandon the rest of the gene pool in which they find themselves, then it may prevail. Otherwise the majority are likely to win out over any innovative improvements in the genetic system. Sex could thus be described as a heavy subsequently constructed structure, built on top of the genetic mechanism which tends to fix it in position.

The above description applies to small, gradual changes in the genetic structure, but is not relevant to theories of genetic takeover of the kind described by A. G. Cairns-Smith, which remain possible for as long as a superior genetic substrate can be envisaged.

Having mentioned Cairns-Smith's work, another of his metaphors seems appropriate to the discussion. He describes the lock in phenomena concisely, using the metaphor of building a stone arch. This is a classic metaphor illustrating that the initial stages of final constructions may be best seen by envisaging preliminary supporting structures; in the example of the stone arch, this may be a mound of earth. If the mound of earth is then removed, the arch becomes suddenly brittle and locked in to a state where all the stones depend on each other. Once this dropping away of preliminary structures has occurred, there is often no going back, even if the resulting construction contains aspects which are no longer appropriate to the uses to which it is being put.

It seems unfortunate that Cairns-Smith's work has not found a wider audience within the A-life community. One problem may be that a well known A-life enthusiast Stuart Kauffman has proposed a conflicting theory of the origin of life to the one Cairns-Smith proposes, involving autocatalytic sets. The author believes that this barrier may eventually be resolved by problems within the autocatalytic theory. Cairns-Smith's own critique of models of chemical evolution seems to be applicable, as autocatalytic sets would seem to have what Cairns-Smith describes as a low ceiling in terms of information storage. This, if true, would prevent them from being considered seriously as candidates for the first genetic material.

Turning back to chromosomes, meiosis will now be examined. There seems to be some control by the chromosomes over the details of how meiosis operates. For example, some parts of the genome are more sexy than others. Different parts of chromosomes also have different baseline mutation rates. However, these effects provide only a very minor structural element to meiosis. Apart from the near immunity of that genetic wasteland, the Y chromosome, there are no parts of the genome which are treated with additional respect by the process.

Meiosis' lack of discrimination may be an aspect of its even-handedness. If the meiotic process can be influenced by a particular gene in a manner that makes that gene more likely than its alles to find its way into gametes, then that gene will come to dominate, even it it has other severe side-effects on individuals who bear it. This kind of gene is often referred to as an outlaw gene. There is subsequent pressure on meiosis from the other genes in the organism which all suffer from the adverse effects of outlaw genes. It has been theorised that under some conditions the outlaws are fairly effectively suppressed. It may be that the form of the suppression forces meiosis to be simple, as any structural complexity offers outlaws opportunities for subverting the process to their own ends.

The some conditions under which the outlaws are suppressed are theorised to be as follows:

Outlaws that act by subverting meiosis or around the point of conception are known as segregation distorters. They may act at other points in the lifecycle of the organism, as in the example of a behaviour pattern, influenced by a gene on the Y-chromosome, causing fathers to invest more resources in bringing up sons than daughters.

Outlaw genes not only help themselves to a larger slice of the genetic pie than they would normally receive, they usually help their chromosomal neighbours too. Sometimes (depending on how the outlaw pursues its aims) the entire chromosome containing the gene is helped. If an organism has a large number of chromosomes then a larger proportion of the organism's genes suffer as a result of their association with the outlaw than would if the organism had fewer chromosomes. These are more likely to act in unison to suppress the effects of the outlaw genes, than if there are, for example, only two chromosome pairs, in which case one of them may want the outlaw to succeed because its genes will directly be levered into the next generation by the outlaw's mechanism. It will try to block any suppression effects which the other chromosome pair would like to impose on the outlaw.

If this reasoning is correct then it follows that multiple chromosomes and sex go together. No suggestion is being made here as to which of these adaptations came first. There may be other pressures favouring multiple chromosomes (including structural and engineering considerations). However, organisms using sex with a low chromosome count may need some variations in its style of meiotic division if they are not to become prey to outlaw genes.

The reason for interest here in the structure of chromosomes is as follows. Genes display associations with other genes in the genome, in that some groups of genes are more likely than others to find themselves sharing bodies in offspring. These genes are described as being linked to one another.

The mechanism behind linkage is very simple: genes that are on the same chromosome are linked roughly in proportion to the distance they are from one another along it. This distance is related to the probability that meiosis or mutation will act to cleave the two genes involved. Genes on different chromosomes are considered not to be linked at all - their probabilities of making their way into a body in the next generation are normally not connected at all.

Linkage is important. Even in asexual organisms where rapid genetic shuffling is not the norm, it may play some role. Characters which are developmentally related or inter-dependant would gravitate towards close linkage. Genes that need one another for their mutual well being would be inclined to find themselves next door to one another. Within linkage groups, genes would tend to cooperate more with their neighbours than with the rest of their particular genome when they encounter copies of them in other bodies. Recognising a copy of your close neighbour out in the world may be exactly equivalent to recognising a copy of yourself.

The means by which genes are brought together is of course natural selection, but of a slightly unusual type. Usually selection may be thought of acting on the phenotypes of organisms, but in any individual organism genetic linkage displays no obvious signs of its effects. It is only when the organism reproduces that some important groups may be separated if they are not closely linked. The selection pressures may be thought of as being towards evolvability, and away from a tendency towards rapid genetic decay in descendants. The recognition of the existence of such selection pressures is not yet widespread. For a justification of the plausability of such selection pressures readers are hereby referred to the chapter titled Kaleidoscopic Embryos in Richard Dawkins' recent book, Climbing Mount Improbable.

Linkage comes to follow and mirror the structure of the organism. This is not meant literally - in the sense that genes associated with different adult bodily parts would like to form an isomorphic structure on the genome - but giving due consideration to the developmental history of the organism. In the complex and mysterious growth processes of embryology, genes can be seen as waltzing with other genes, and their partners tend to become their eternal soulmates.

The main point to be made here is that distance is a very crude one-dimensional symmetrical measure. If an organism could have multi-dimensional asymmetrical, conditional, with complex interdependencies as relationships portrayed by links, then it would benefit in that the complex relationships and dependencies between genes in embryological processes could be captured more clearly.

In the additional organism, genes may ony be more or less strongly linked to to one another. It would seem desirable to be able to represent more complex relationships between genes. For example, the relationshoip: "Gene A helps Gene B if in the presence of Gene C, but otherwise it hinders it", is not a relationship which can be easily expressed in terms of simple linkage. Such relationships between genes most certainly exist and have been tracked in the studies of genetically inherited diseases.

Tree-shaped chromosomes are referred to here. This is for a number of reasons. Heirarchical trees are familiar data structures to many. They are easy to implement on existing computers. A simple representation can be imagined as a one-dimensional string of characters using nested brackets to indicate the branching depth.

There is no need to imagine anything as complex as branching in the physical DNA used to encode genes. A simple one-dimensional structure may be used, but it should contain markers (serving the same function as the brackets referred to above) which encode information relating to the tree structure, and which influence meiosis. It is easy to imagine the sexual process as being represented by a pruning and grafting process between chromosomes.

I do not claim that tree-shaped is in any sense an optimal structure and it may be preferable to consider genes as being related to groups of linkage properties in an orthodox object-oriented manner.

There seem to be two main problems with the idea as it has been presented.

Bearing in mind the existing state of the genetic substrate, the advantage of having a better map between genetic and developmental/phenotypic structure may be counterbalanced by a corresponding high cost of implementation. The existing genome is one-dimensional and complex links would probably have to be represented physically on it in some manner, and then read in the process of meiosis. I have mentioned above that the redundancy of the existing genetic code could provide one way of coding such information. Another method (which I have been advised is already used) would be to provide marker strings of DNA which act as cutting marks for particular enzymes.

Also, this kind of selective meiosis would be just the kind of mechanism outlaw genes would love to get their hands on in pursuit of their own subversive goals. This would have to be guarded against. In the same way that splitting the genome into unlinked independant sections may help suppress the action of outlaw genes in the existing system, exactly the same strategy could reasonably be expected to reduce their effects in an organism with tree-shaped chromosomes.

Another possibility might be to store information about structure and linkage non-locally. This would certainly prevent any linkage effects between the markers and the genes which they are marking, but it is no longer clear how this linkage information could then be inherited with the genes which they are marking. Without this, the linkage information would not be useful.

Between them, the two disadvantages mentioned above may contribute to the apparent absence of such geometries from modern organisms.

Naturally the two problems described above apply mainly to small variations on the existing system. Messianic A-life heads - who feel as though they have an opportunity to start from scratch - may wish to avoid constraining themselves by such considerations.

tim@tt1.org | http://texturegarden.com/