log of research in progress 2001

Nov 16th 2001
I have finally settled on a way to introduce the thesis work that feels satisfying - i.e. covers both EC and EB in a way that doesnt overstate either. I decided to make crossover and symbiotic encapsulation subsets of 'compositional evolution'. That way the results on crossover are not just background leading up to the real thing (encapsulation) but are an integral part of the results. Then compositional evolution - the exchange of preadapted subsystems between parallel lineages - is contrasted with 'accretive evolution' (adding small changes gradually). This leaves out mechanisms of large genetic change from gene-duplication (within a lineage) and large phenotypic change from ontogeneis (also within a lineage) - but Im not going to look at those.

Chapter writing furiously underway, with good progress so far.

May 9th 2001
...big sigh... I cant believe its May already!
Well, I guess I'm not graduating in May then... Im aiming for Dec 2001 now.
Im looking for postdocs starting in spring 2002.

I feel like I havnt achieved very much in the last six months - but really the problem is that I havnt acheived the things I thought I would.
Specifically, I havnt done the 'soft joins' version of SEAM, so I havnt applied SEAM to anything practical.
And, I havnt got very far with writing a thesis framework that supports the CS view and Evolutionationary Biology (EB) view.

But, what I have done is got quite a long way with describing SEAM in a manner which makes it accessible to an EB audience. The trouble was that, although its easy to describe SEAM as a form of genetic algorithm when talking to a CS audience, its not so easy to describe what SEAM is when you want to realte it to real biology. Partly this is because of a need to related to existing biological terminology, concepts and models that Im not (was not) familiar with. But a deeper problem arises simply from the fact tht in CS its OK to do whatever it takes to make something that works - and 'what you really need' for effective adaptation, is not necessarily 'what you really have' for adaptation in natural systems. So, I have to move from explanatory language like 'it has this mechanism because otherwise you'd get stuck in a local optima' to things like 'it has this mechanism because thats a reasonable approximation to this known situation in natural systems' - quite a different form of 'explanation'.

Anyway, I've made quite a bit of progress with that I think. I've learned a lot about things like Wright's Shifting Balance Theory, Panmictic (freely-mixed/unstructured) populations and sub-divided populations, mechanisms for selection on sets of alleles rather than individual alleles, mechanisms for symbiotic union like horizontal transfer, allopolyploidy, shared vertical transmission, and other mechanisms of co-dispersal, and concepts of the niche and competitive exclusion. Just as importantly, Im beginning to learn the kinds of questions that biologists want answered. They dont ask 'how does it need to work to get the optimal configuration of features', they ask things like 'given these conditions and assumptions, what will happen to the distribution of alleles in the population'. For example, whats the probability of fixation of an allele in a panmictic population and in a structured population.

There are some 'gaps' in the general EB framework, as I understand it so far, where I have something to offer.

1) Poly-loci epistasis models.There aren't any models of hierarchically structure epistasis. There's lots of models of two-loci epistasis. And some models of randomly-structured interaction between many loci (well, only NKs really). But there arent any models where the epistatc interactions between small sets of alleles are similar in nature to the interaction between large sets of alleles - i.e. there are no scale invariant epstasis models.

2) Integration of ecosystem models with population models. Wallace Arthur makes a point of the fact that many models of conditions for stable coexistence of polymorphisms are applicable to conditions for stable coexistence of species in an ecosystem, and vice versa. But no-one seems to have produced a model that integrates population dynamics with ecosystem dynamics and importantly, shown how they might interact with one another. Its a very difficult thing to do - lots of assumptions need to be made and pinned down. But there is a key insight in treating intergenomic epistasis/competition/coexistence and intragenomic epistasis/competition/coexistence with a single integrated model of interaction between 'entities'.

3) In light of (2), it makes sense to talk about processes that unite genomes together into larger composites. The symbiotic union of two different species into a single composite entity can be seen as the 'canalisation' of favourable intergenomic interactions (making them into favourable intragenomic interactions).

4) Abstracting away population dynamics and ecosystem configurations with the use of concepts like 'could there be an arrangement of the entities in the ecosystem such that these two entities could coexist' => Pareto dominance.

5) With 1 2 3 and 4 together we can start to build an integrated adaptational model of the major transitions where interspecific competition, the formation of new entities from the composition of simpler entities, and a bunch of other previously disconnected phenomena are brought together to explain their adaptive utility.

See publications for further background.