I've been asked to attend an upcoming workshop on modeling the evolution of aging. Modeling in this sense means building a computer model that provides insight into a process. Most of the people who will be there have some experience in modeling, which I don't. So I decided my roll, as the participant most focused on evolutionary theory, should be to make sure that they are doing evolution right. We were asked to each produce a seed, or opening idea, to get the conversation going. Following is a draft of mine. Note that while you can figure out what most of this means, it was written for other scientists and so includes a smattering of jargon and techniquese.
As an evolutionary biologist, I would like to encourage those modeling the evolution of aging to include in their models actual evolution. Optimization models and even Markov Chain models, while very useful, are not evolutionary. By this I mean that they do not include populations changing through descent with modification. To meet a biologist's definition of evolution, a process must include individuals who are reproducing and the offspring must be modified copies of the parents. This requires a population of individuals with heritable traits and mutation rates which modify the parents' traits in the offspring. In order for adaptive evolution to occur, these heritable traits must also influence how many copies of its genome each individual passes on to the next generation.
Markov-chain models, while somewhat closer to evolutionary, still lack the aspect of a population, which is essential for evolution. In many cases the outcome of evolution will depend on having competing or interacting sets of genes within the same population. This cannot be meaningfully understood if the whole population is assumed to have only one set of genes at any one time.
A truly evolutionary model of aging must therefore be fairly complex. It must simulate individuals, who have age-specific mortality and fertility probabilities. These age-specific schedules must be determined by a set of genes. These genes in turn must be determined by a process of inheritance and a process of mutation.
Using such a method, we can address questions that are difficult to get at through optimization. For example, suppose we would like to know why closely related populations have similar patterns of aging, even when they live in different habitats, or occupy different niches. This pattern has been observed in comparative data and comes under the heading of phylogenetic inertia. With an evolutionary simulation, we can impose environments which mediate the relationship between the genes and the demography. We can then ask what characteristics of the environment or what characteristics of the relationship between environment and demography would allow the starting point (that is the initial genes and demography of the population), to influence the ending point (that is the genes and demography the population ends up with).
To take another example, optimization models generally lack any information on the structure of the genome or the process by which that genome changes. However, genomic structure and mutation process are not irrelevant to what demography the population evolves. An evolutionary simulation will allow for modification of the genomic structure or the mutations process. Compare for example, two populations, each of which has a genetically controlled pattern of investment in various tasks such as reproduction, repair, growth or immune function. In population A, as many genes control this at the beginning of life as at the end. In population B, many genes control the pattern of investment early in life, while relatively few are still influencing late life investment. In both populations the genes affecting these investments are subject to mutational pressure and to selection. In each a mutation-selection balance will emerge, but these mutation selection balances will differ between the two populations. The two populations living in the same environment will arrive at different demographies, each nonoptimal.
These are but two of the many complicating factors which can be explored using an evolutionary simulation and are difficult to get at in a model that does not include explicit evolution. Of course models should be simple enough that one can figure out what factor is influencing what outcome. A model cannot include every complicating factor biologists might like to throw in. As such, I propose a modular evolutionary simulation. By this I mean we start with as simple a model as we can which still has real evolution going on and we write it in such a way that one can add more complicated processes. For example, the basic model could have an extremely simple process of mutation, but could be written such that that this process is easy to remove and replace with a more complicated mutational process. Reproduction could be clonal, but again that process of inheritance could be coded such that it could be pulled out and replaced with sexual reproduction by someone who is interested in what effect the mode of reproduction would have on the evolved demography. The environment could be extremely simple and replaceable with a more complicated environment. I am a slow and inexpert programmer but I imagine that it would not be impossible to write such a simulation in a way that genome, inheritance, mutation, environments, and demography are interacting pieces which can be replaced as one replaces the batteries, bulb, wire, switch and casing of a flashlight. One need not modify the casing to replace acid batteries with rechargeables, or replace rechargeables batteries with lithium rechargeables. One can swap a white bulb for a yellow one without modifying the wires or switch. A properly designed base simulation would allow each of us to experiment and still be able to compare our results without any one model becoming unnecessarily complex.