According to the theory of evolution, why do we die?

Here is an interesting answer to that question found on Quora.
It makes me wonder if/when dying is an evolutionary advantage. I am going to think about an experiment on the matter.

Answer by Suzanne Sadedin:

Excellent question. And before I explain the real answer, which is rather mind-bending, here are some previous arguments and why they are wrong.

Myth 1: We die to make room for younger generations.
Genes are selfish, and each individual body is a vehicle for a collection of genes. These genes are selected to favor the survival of copies of themselves. Since parents and offspring use the same resources, the death of a parent creates room ecologically for just one offspring. Each gene in the parent has a 50% chance of appearing in this offspring. But it has a 100% chance of appearing in the parent, because it’s already there. It’s never, then, in the evolutionary interests of a parent to die so an offspring can replace it.

Myth 2: We die because our cells/DNA get damaged with age.
This like saying bad drivers die because of blood loss. It’s a proximate mechanism of death, not the evolutionary cause of mortality.

Our somatic cells (the cells that are part of our body) do indeed suffer occasional mutations as they divide. These mutations can kill or damage cells, which is annoying but not generally a big problem as we can make more. However, the worst mutations do something much more dangerous: they help cells to survive and proliferate. That’s how you get cancer. Because this risk accumulates over time, cells are normally allowed only a limited number of divisions before they undergo cellular senescence, that is, they die. But the genes that cause cellular senescence can also stop working. So that’s one of the ways in which we get old: our somatic cell lineages get older, damaged and mutated, and some become cancerous.

However, the cell/DNA damage idea assumes that this isn’t something evolution can counteract. And that’s clearly false. Lifespan and cancer rates differ between species, and not in the ways you would expect if they were determined by cell/DNA damage. For instance, once you take into account body size and phylogeny, DNA repair doesn’t correlate with lifespan. Lifespan does, however, correlate with ecology: mammal species who typically lead risky lives die younger (even if you protect them from those risks). At one extreme, in the harsh Australian bush we find the male agile antechinus, who dies of stress at the end of a single breeding season. At the other extreme, the naked mole rat can live for three decades in its peaceful underground colonies.

This gets even more puzzling when you start to look at genomics. We have a whole suite of genes devoted to keeping our genome pristine. My favorite is a clever gene called P53 that acts as a “gatekeeper” for cell division. If the cell has too many mutations, P53 will halt division and activate repair mechanisms. If that doesn’t fix things, it will make the cell commit suicide. Mutations that break P53 are involved in about half of all human cancers. Now, here’s the rub: there’s a whole family of genes related to P53 in other mammals, and some work better than others. Naked mole rats, as it happens, have two particularly awesome versions that completely protect them against cancer.

We also know that it’s perfectly feasible for genetic modification to immortalize cell lineages, and that going through a haploid stage is not essential for maintaining cell viability. How do we know this? From the strange case of the 11,000 year old dog. The dog as an individual is long dead, but her cells survive today as an infectious cancer on other dogs’ genitalia. There’s also a quaking aspen in Utah whose roots are at least 80,000 years old.

The same applies to permanent organ damage. Some organs heal and regenerate, some don’t. Some species can regenerate organs that others can’t. A salamander can grow a whole new leg. There’s even a jellyfish that can reverse its development when it’s damaged. All in all, natural selection is clearly capable of creating creatures who can fix cellular and DNA damage and repair damaged organs.

So: evolution can fix these problems for us, and it doesn’t. What the heck, evolution, aren’t we friends?

Well, no, actually, evolution is not our friend. If anything, it’s our genes’ friend. And there’s a very good reason our genes don’t actually care about us.

Mutations are a problem evolution can fix. But death isn’t. Accidents happen. Diseases happen. Sabre-toothed cats happen (well, not any more, but you get the point). No matter how hard our genes try to help us survive, sometimes they’re going to fail. These failures are often, as far as your genes are concerned, random. And that means our genes can’t afford to get too invested in the survival of any individual. In the long term, the only way a gene can survive is to spread — to copy itself through a population.

So from a gene’s-eye view, every investment in your survival is a potential trade-off with the creation and survival of your potential descendants. And, rather obviously, the more likely you are to die randomly, the less it makes sense for your genes to invest in the survival side of the equation.

Every day of your life, the Universe in effect rolls a pair of many-sided dice. Snake eyes, you’re dead. Every day the probability that the Universe has at some point in the past killed you increases. And at some time after your birth, on average, you’re dead.

Look at this from your genes’ perspective. Your genes don’t know about you specifically, their behavior is selected based on statistics. They don’t want to invest in somebody who is, on average, dead. Younger people are, on average, more likely to be alive. So if your genes have to choose between investing in (on average) the survival and/or reproduction of a young person versus an old one, they’ll pick the younger one.

And quite often they do have to choose. Early in development, for instance, you really need genes that allow lots of cellular proliferation. Your body can’t grow without it. But too much cellular proliferation when you’re fully-grown is a big problem. So it’s a delicate balance, and what’s good for you when you’re a kid can be bad for you when you’re grown up. There are other genes that manage these risks by switching genes on and off throughout your life, but that makes the network even more complex and failure-prone. You end up with an intricate genomic dance going on throughout your whole life. So it’s hardly surprising that some genes end up helping you now and harming you later.

One example may be Huntington’s Disease, a horrible dominant genetic disorder that slowly destroys your brain and kills you. The disease usually starts to affect people in middle age. However, young people with the Huntington’s gene have more children on average. It’s thought that the Huntington’s gene strengthens the immune system by increasing activity of P53, making them healthier and more fertile. Other possible examples include atherosclerosis, sarcopenia, prostate hypertrophy, osteoporosis, carcinoma and Alzheimer’s disease.

As life goes on, your genes effectively stop caring what happens to you. After a certain point, it’s so unlikely that you’re still alive that your genes can safely assume you’ll already be dead. So your genomic programming can contain all sorts of wacky stuff that only kicks in after this point, just because there’s no noticeable selection against it.

The really fascinating part (by which I mean the really depressing part) is how this effect reinforces itself. The more likely it is that you’re dead, the less your genes care about you. The less your genes care about you, the more likely it is that you’re dead. And this has been going on throughout our evolutionary history, so we’ve accumulated all sorts of weird malfunctions that kick in late in our lives. The human genome is riddled with them, and most of the genes involved are also part of normal development and reproduction. These malfunctions cluster around a certain age: the age when evolution stops caring about us because, statistically speaking, we’re already dead.

So mortality is an evolutionary prophecy that fulfills itself in a multitude of ways. And that’s why there’s no single key to eternal life. Poor old Gilgamesh.

A review of the evolutionary theory of aging
The evolutionary genetics of ageing and longevity
Huntington’s Disease: http://www.sciencedaily.com/rele…
More possible examples of antagonistic pleiotropy in human diseases:
A Darwinian-evolutionary concept of age-related diseases
Antagonistic pleiotropy as a widespread mechanism for the maintenance of polymorphic disease alleles
The 11,000 year old dog
Page on sciencedaily.com
The 80,000 year old quaking aspen
National Park Service
The jellyfish that reverses its development
The Immortal Jellyfish
Aging in different mammal species is related to environmental vulnerability:
Mammalian Aging, Metabolism, and Ecology: Evidence From the Bats and Marsupials
Large-brained mammals live longer
Longevity in mollusc species correlates with life history traits:
Maximum Shell Size, Growth Rate, and Maturation Age Correlate With Longevity in Bivalve Molluscs
DNA repair doesn’t explain longevity of species (when body size is factored in):
DNA Repair and the Evolution of Longevity: A Critical Analysis
Anti-oxidant activity also doesn’t explain longevity differences between species:
Antioxidant enzyme activities are not broadly correlated with longevity in 14 vertebrate endotherm species

According to the theory of evolution, why do we die?

On how to live longer (and being an adult)

Hi!
In this post I’m going to show you what happens when our sheep and wolves are born directly adults and, on the contrary, when a period of childhood is defined. You might want to read the first post before, it is very short and explanatory.

You can now install the Game of Evolution very easily from the GitLab page of the project.

There are two parameters you can play with: sheepA and wolfA. They are ranges (e.g. 20-80) which define from what age (in number of turns) to what age a given animal can procreate. For example, if sheepA = 20-80, only a sheep with age between 20 and 80 will be able to procreate; if a sheep that is too young or too old tries anyway, nothing happens, it just loses its turn while he could have moved or eaten some grass.

In a first experiment, let’s see what happens when both sheep and wolves are born adults.
For now, the maximum age for sheeps and wolves is set to 100 and 200 turns respectively, which means that if a sheep is still alive at the beginning of its 100th turn, it will die instantly, no matter its health or hunger.
So, here we want sheepA = 0-100 and wolfA = 0-200.
The full configuration file I have used (with the –configFile option) for the first experiment is here. This generates an file (named demographics-a.txt according to the DemLogger option) with some demographic information that you can analyse with the demographic_analysis.nb Mathematica notebook in the mathematica/ folder of the project. This requires Mathematica, but if you don’t want to install it you can still use this cloud notebook I made (I don’t know how long it is going to be online, though).
Here are the evolutions of the population size and the life expectancy of our animals (on x-coordinate is the number of turns):

Experiment a: both sheep and wolves are directly born adults

We observe a transitional phase at the beginning, which is perfectly normal: during each turn one sheep and one wolf with random DNA are generated and put on the map. Those random animals are very unlikely to act “rationally” — that is to eat when they are hungry or to move when no food is nearby for instance — and thus die very quickly. The population stays very small until a smarter sheep appears, an sheep which knows what to do to survive and procreate (note that there is no sexual reproduction mechanism in the Game of Evolution yet, what happens is more like a mitosis). This lead to an explosion of the sheep population, which is stopped by the limit of the map: the map has 64*128 tiles, each on them producing one unit of grass per turn while a sheep needs an average 4 units of grass per turn (so the map can sustain a population of 2048 sheep maximum). After that, the wolves can thrive.

The average values (after time t = 50 000 in order to skip the preliminary phase) are:

• sheep population: 730
• wolves population: 660
• sheep life expectancy: 5.1 turns
• wolves life expectancy: 23 turns

Experiment a on a longer period of time

We clearly see that the animals die very young (they could theoretically live 100 and 200 turns respectively).
On the graph it looks like the wolves life expectancy is increasing, so we can wonder whether after a time long enough it will reach the theoretical limit of 200 turns. The answer is no: I have run the same experiment for 2 million turns and here are the results:

The average wolves life expectancy is 24 turns.

So our animals are kind of dumb. In the game of evolution living longer can only be an advantage (because you can reproduce more), but in this simulated universe it doesn’t seem strong enough to lead to a significant evolutionary pressure.

As second experiment, let’s see what happens if the sheep can only procreate after their 10th turn.
We can set sheepA = 10-80 as in the configuration file I have used. Notice that 10 > 5.1, the sheep life expectancy in the previous experiment (I’m realising now that in addition to the life expectancy, that is the average age of death, the median age of death would have also been interesting data).

Experiment b: wolves are born adults but sheep can procreate only from age 10

The average values (after time t = 50 000 in order to skip the preliminary phase) are:

• sheep population: 740
• wolves population: 290
• sheep life expectancy: 17 turns
• wolves life expectancy: 21 turns

As we can see, the sheep that have developed in this world live much longer. We can also observe that although the characteristics of the wolves have not changed in any way, their population is much smaller, while the sheeps population size is the same as previously. It might be relevant here to keep in mind that for now the wolves can’t tell the difference between two sheep, in particular between a young one and an old one.
Another thing is that the wolf population seems more unstable. We can see a few drops in its size and the wolves life expectancy

Third experiment: the sheep are born adults but the wolves need to be 20 to procreate.
configuration file

Experiment c: sheep are born adults but wolves can procreate only from age 20

The average values (after time t = 50 000 in order to skip the preliminary phase) are:

• sheep population: 1100
• wolves population: 530
• sheep life expectancy: 6.6 turns
• wolves life expectancy: 25 turns

Compared to experiment a, everyone live slightly longer, there are more sheep and less wolves. The evolution of the sheep population is funny, I’m not sure what to say about it (if you have any idea of what kind of statistical analysis I should perform, don’t hesitate to leave a comment — this is generally true, if you have any comment or criticism, please tell me).

Fourth experiment: sheep are adults at 10, and wolves at 20.
configuration file

Experiment d: sheep are adults at 10 and wolves at 20

The average values (after time t = 50 000 in order to skip the preliminary phase) are:

• sheep population: 920
• wolves population: 230
• sheep life expectancy: 25 turns
• wolves life expectancy: 33 turns

In this settings life expectancies are very high! They are still far from the biological limits (100 for sheep and 200 for wolves), but still much higher than initially. As when only sheep need to be adult to have offspring (experiment b), the wolves population is very low, and the system seems unstable: look at those events at time t = 130000 and t = 150000. I don’t exactly know what happened there, but it is not surprising that because we have made the universe more complex, putting constraints on its inhabitants, some behaviours that were fine before now lead to demographic disasters.

I am now finishing this post, I hope you found it interesting. If you have any thoughts about all that, feel free to share them here. I would be very glad to discuss these matters with anyone, especially if you have ideas about what to look for or what kind of changes to make in the simulation. And don’t forget that you can get the program for free here; it is an open source project, you have full access to the code if you want to look at it (it’s still a bit messy but I’m cleaning it gradually) or even improve it!

For my next post, I am considering showing how allowing the sheep to defend themselves lead to a gregarious instinct (I won’t hide you that one of my main goals with the Game of Evolution is to observe complex social interactions).

Hello world!

Hi!

I’m starting this blog to share my discoveries in the Game of Evolution.

The Game of Evolution is a computer program I am developing to simulate a (very) basic world and its inhabitants – currently sheep and wolves. The fun part is each of these animals has DNA that is passed on to its children (if there are any) with slight random mutations (the first animals are created with random DNA). Because the DNA determines the animal’s behaviour and thus its ability to survive and procreate, it is subject to a strong selection process, very similar to real life Darwinism!

My aim is to observe complex and “intelligent” behaviours. Of course, this is probably not easy, due to the fact thatsuch behaviour must simultaneously represent actual evolutionary advantages in the simulated world and be cognitively accessible to the animals. The latter relates to the cognitive models used to compute what actions are chosen by each animals at every time step.

At every time step, each animal is asked to chose between a fixed set of actions (like going up or left, doing nothing, having a baby, etc.). This is done by an artificial neural network (described by the DNA), whose input is a “perception vector” representing what information about the current state of the world the animal has access to. So the cognitive power of an animal depends on its perception vector (How far can it see? Can it tell the difference between two animals? What does it know about itself? And so on.) and on the structure of its neural network (size and depth of the network, recursivity, etc.).

In short, the goal will be to find some rules of the game, i.e. the physics of the world and the nature of its inhabitants, so that these evolve (or not) to acquire interesting social and individual characteristics. I already have witnessed a few things, but this will be the subject of subsequent posts. 🙂

I will very shortly make the whole code accessible (I just started a GitLab repository). It should be very easy to install and run the program. EDIT: It’s here!

See you later!