I have to say, while the UD “newsdesk” is terrible source for comment on scientific news, the links themselves are often interesting. Today, the UD “newsdesk” reports on a pretty interesting study, reported in Nature, here, and a preprint of which seems to be open access here
It’s been apparent for a while from Genome-Wide Association Studies (GWAS) that “risk” alleles for various mental disorders, despite being statistically significant, have extremely small effect sizes. In other words, while the studies show that many mental disorders are indeed associated with specific alleles (and we already know that many are highly heritable, including schizophrenia, ADHD and Alzheimer’s), there aren’t just a few rogue alleles of large effect (well, there are, but they are far rarer than these disorders), but instead, a whole cocktail of alleles with very slightly raised Odds Ratios for certain disorders (and some are shared between multiple disorders). This means that the vast majority of people carrying these “risk alleles” are perfectly fine. That would help explain why they have not been weeded out by selection.
But the interesting thing about this article is that it provides evidence that some have been positively selected for – that they confer an advantage. Moreoever, they are alleles of genes implicated in brain function.
We’ve known for a while that the “E4” variant of ApoE (carried by 1 in 7 of us) is a risk factor for Alzheimer’s, and that it is also associated with enhanced cognitive performance early in life. So it’s really interesting to see an actual systematic study of the evolutionary history of other genes implicated in brain function in general, and their downsides.
And, Denyse: it’s only “Darwinbabble” if you don’t speak it.
On the UD thread, Mung wants to know
and asks:
We can actually set this up as a computer model.
And the answer is that the “selective value” of the variant possessed by each individual affects the probability that it will leave offspring.
To keep it simple, let’s set up a population of simple non-sexually reproducing critters. In each generation they all produce one offspring, after which, approximately half the population then dies.
Variants with a “neutral” selection coefficient will have a .5 probability of dying in any one year. Those with a “positive” selection coefficient will have a probability of less than .5 of dying; those with a “negative” selection coefficient will have a probability of more than .5 of dying.
So the answer to your question is that you neither “perform the random sampling first” nor “look at the selective value first”. You simply randomly sample from the non-flat probability distribution. Sometimes you will kill a critter with low probability of being killed, but far less often than you will kill a critter with a high probability of being killed. And for the neutral ones, it will be a coin toss.
And the way you tell whether there is selection going on when you observe a particular variant reducing or increasing in prevalence, is by doing a simple null hypothesis test. You work out the probability of a variant reducing or increasing its prevalence under the null hypothesis that it is neutral (in this case, a .5 chance of surviving each year) by the amount or more that your observed variant reduced or increased, and if that is very unlikely under the null, you conclude that selection was operating, p<.00whatever.
By the way, this should also explain why saying that “selection can stop things getting worse, not make things better” is wrong. Selection can be both positive and negative – a variant can increase its bearer’s chances of surviving each year relative to the average (or relative to its parent’s) or reduce it. Both are perfectly possible.
How I actually do it, usually, on a computer, is, for each critter, each generation, I draw a random number from a flat distribution between 0 and 1.
If that critter’s probability of surviving (i.e. what I assigned to it as it’s “selection value”, or, in practice, its “kill value” is more than the random number I drew, I let it survive. If it is less, I kill it.
For critters with a “kill value” of .5, it will be a coin toss – equiprobable that they live or die. For critters with a higher kill value, there’s much higher probability that my random number will be less than the kill value, so that implements the higher probability of death. For critters with a low kill value, it’s only rare that my random number will be less than that, so they have a good chance of survival.
So it’s completely stochastic – the selection is always random. It’s just not equiprobable. And variants with low “kill values” will tend to become more prevalent, while variants with “high kill values” will become less so. All, however, will be subject to “drift”.
And, as I said, we can compute the probability of a given amount of prevalence change under the null of “neutral” and, if that is low for the observed amount of prevalence changed, we can conclude “selection”.
Science is Greek to Dynese.
Seriously though, has any follower of kariosfocus ever considered the implications of alleles on his isolated islands metaphor?
Look more at the comments on the UD thread, it’s interesting how widespread is the misunderstanding of “random” – for many it seems to mean “equiprobable”.
That’s a recurring error in a lot of ID writing, and it’s even there subliminally in Dembski’s.
Not that Dawkins is exactly sound on the topic.
ID advocates and creationists do seem to have trouble thinking about sampling from distributioms where there are unequal probabilities. They can’t see that the unequal probabilities mean that natural selection is acting.
If we have 1000 parents with unequal feritlities f1, f2, f3, … f1000 we sum up these 1000 numbers, call that F, and then consider the probabilities f1/F, f2/F, f3/F, … f1000/F. These add up to 1. Now all you need to do is draw a parent at random with these probabilities and we have both random genetic drift acting, and natural selection acting.
There are reasonably fast ways to do this. But somehow our friends at UD and their compatriots here just can’t understand this process.
I’d like to thank Elizabeth and Joe for their explanations.
So how do we tell which result is an effect of [due to] random genetic drift [as a cause] and which result is an effect of [due to] natural selection [as a cause]?
This was the original question raised in that thread and afiak it still remains unanswered.
24:35
If we want to invoke natural selection ever we need to have a null hypothesis that we can reject … we need to know what the neutral expectation is.
: 27:40
It all comes down to one parameter, but the problem is there is a lot of biology in this one parameter. Mutation is in there drift is in there and selection is in there.
Care to comment on Lynch’s statements?
Mung,
If you have a slightly unfair coin, one which comes up heads 52% of the time, how do you decide whether in a particular toss the result “heads” is due to the 2% bias or to pure chance (or maybe even to some delibarate manipulation on the tosser’s part) “as a cause”? Who says a phenomenon has to have a single unique cause (rather than being contingent on lots of different factors)?
Well, as I said, you do a null hypothesis test.
You figure out the probability of a given allele increasing by the observed proportion or more if the variant is neutral (i.e. has the same chance of leaving offspring as the average), and if that probability is very low, you conclude that selection is probably operating.
It’s exactly the same process as detecting whether a coin is fair: if you toss the coin 500 times (to pick a popular ID number 🙂 ) and it comes down heads 60% of the time, when you expect 50% under the null hypothesis that heads and tails are equally probable), then you can compute probability of that outcome under the null, using the binomial theorem (infinitessimal in this case).
And the point is that “selection” and “drift” aren’t two separate forces, like gravity and magnetism, say. They are metaphors two aspects of a single process: the renewal of a population through birth and death. When the chances of leaving offspring before they die are completely independent of genotype, there is no genetic bias on this process; the probability distribution is flat. We call this “no selection”. Where there is a genetic bias – when some genotypes have a better chance of leaving offspring before they die, we call this “selection”. But all it means is that the probability distribution is not flat with respect to genotype.
This is why lots of people (e.g. me) get annoyed when Dawkins says “mutation is random but natural selection is non-random”. I know what he means but it’s very misleading.
“Selection” is random – stochastic. It is simply a bias on the probability distribution of reproduction, with respect to genotype – a non-flat distribution. Like a die in which some faces are more probable than others; i.e. not a fair die. But still “random”.
Mutation is the same – some mutations, and some kinds of mutations, are much more probable than others, and in particular, mutations have little or no effect on the chances of leaving offspring are far more probable (occur more often) than mutations that have a large effect. Again, it’s a non-flat probability distribution.
What Dawkins is trying ineptly to say is that the chances of a given beneficial mutation occurring i.e. one that increases success in the current environment is is not increased because the current environment makes it beneficial, i.e. that they are independent. But it isn’t that one “process” (mutation) is “random” and the other “process” (selection) is not. Both are random processes.
I’d like to add, just for the sake of emphasis, that neither selection nor drift are “causes” in the physical sense. They are aspects of our description of the random resampling process.
*high five*
Also, Lynch says the same thing, but with equations.
Indeed. If I google for definitions, “equiprobable” turns up quite frequently. Would it make sense for everyone who uses “random” to make it clear (at least initially) in what sense they mean “random”. I tend to think of “random” as synonymous with “unpredictable”.
Don’t know whether error is the right word! Omission, maybe! The lack of definitions for key words in ID writings seems endemic. They could start with “intelligent” and continue with “design”.
Is he unsound if he is responding to IDers using “random” in the “equiprobable” sense?
It would be great if we could get Richard Dawkins and Larry Moran to pop in and clarify the issue of selection vs drift.
*clutches lucky penny and wishes*
Well they are on speaking terms and apparently have been in the same room at the same time.
It has far too many meanings to be useful, and people who should no better equivocate without apparently noticing (including Dawkins!)
Some examples:
It is my contention that it is specifically because of the use of the same word “random” to describe the stochastic process by which mutations occur and to describe actions that are unintentional and patterns that are meaningless that the whole idea of Darwinian evolution makes people so frightened, and induces the fear that Darwinian evolution means that WE were unintended, and that our lives are meaningless.
But it means no such thing. It’s just conflation between very different meanings of the word “random”.
Human esigners use stochastic processes to bring about intended and meaningful results all the time.
Why don’t you send them a link 🙂
Let’s give my lucky penny a chance, first. 😉
Sometimes it would be a good idea to use a more precise word to avoid misunderstanding. For example, to call mutations “indifferent” (not happening in order to achieve any particular effect) so as not to create the impression that all mutations are equiprobable (of course they aren’t; there are location-dependent hotspots of variation in the genome).
Larry Moran clearly conceives of selection and drift as two different mechanisms of evolution and gives this as a reason why he is not a Darwinist. [I may have that wrong though, perhaps it’s a reason he gives for calling people IDiots. Maybe both.]
That’s the background (in a nutshell) for the discussion about the relationship between the two, which bled over into the thread Elizabeth found.
Well, I would not characterise it that way. I think it’s much better to see “selection” as biased drift.
Darwin did a great job but his “selection” metaphor has really hampered understanding IMO – it was only supposed to be an analogy with the “selection” of “selective breeding”. There is no “force” called “selection”. There’s just shit that happens, and heritable variance in the ability of organisms to respond to it.
And where a variant makes no difference to the ability its bearer to respond to it, then changes in prevalence in that variant will be completely unbiased. Where there is, it won’t be.
This is welcome to debunk that Alzheimer , probably, is not by inheritance. In fact it happens in old age or when decay comes.
I do not agree there is early enhancement of smarts. Thats a myth there is any aid from our brain to our intelligence.
It leads to a better direction that all these problems are not genetic brain problems but more minor problems with the triggering mechanism for the memory.
There could be iheritance of this but still I say if people didn’t get old this problem would never occur. Its an age issue possibly with inheritance in the triggering problem.
All this is based on great presumptions that are unproven.
These stats mean nothing if selection never made a difference in real human populations with these rare diseases.
how can this disease be selected against when it comes old age anyways? After breeding days.?
Alzheimer’s is certainly highly heritable. But it is true that it generally affects people who are quite old, and thus is becoming more common now that we are living longer.
If it hits when old then it means something.
Either its programmed to hit when old OR , as I see it, it just comes from decay when we get old.
Its like saying wrinkled skin hits people when old but its ONLY vecause we get old.
If people didn’t get old/decay this Alt would not be common.
So I suggest its not a heritable disease at all.
The only thing inherited is a weakness in the triggering mechanism for the memory. i’m not sure that inherited either but if there is a family connection we must conclude something is being given from parents etc to kids.
The thing is I don’t accept there is a ALT disease.
its just a memory problem and so that from the triggering mechanism.
So this is what breaks down in old age. And is inherited.
There is no memory disease being inherited surely.
Its the attributes around it. this recvealed by the fact old age must prompt it.
I jave known people who got it in their 40’s but even thats quite late. I don’t think kids get it. This because its not inherited or a disease. its another memory problem with perhaps triggers being inherited or weaknesses.
I think healing is a better hope if its seen as just a memory issue. Not in the dna.
You are right that it is “the triggering mechanism” that is inherited. What is “triggered” is the production of amyloid proteins in the brain. The disease is about 70% heritable – other, non-inherited, factors can also increase the risk.
It is not like getting wrinkles. I know lots of people who lived past 90 with intact memories.
I know your a researcher and thus thinker on these things, hopefully a dreamer aiming at understanding and healing despite its difficulties, .
Yet saying its amyloid proteins is to me saying nothing about whats going on.
Alcohol can hurt the memory but its not about alcohol but about why the memory is affected.
I say the “proteins” are affecting the triggering mechanism for the memory.
Yet long years of boxing might do the same thing. Being hit in the head moves the proteins too.
I can agree the inherited part is some problem with proteins health but how would one know thats what happened? how 70%? who is examining each case? if its by sampling it could still just be a weakness and not a inheritance in the dna. I suspect if more people lived longer more or all would eventually get this.
So do i. Yet i see in many things that they are said to be inherited diseases but only come in old age. so i suggest its the decay from old age that is the point. no decay and these issues would never be seen.
So i question this inheritance concept.
I retreat to weaknesses in the system being inherited if thats the word.
anyways i’m sure its a triggering problem to the memory regardless of what is or is not inherited.
I say all mental problems are only memory problems. No such thing as brain problems. The brain is just a memory machine. Then our soul is meshed to it.
The brain is not just a “memory machine”, Robert. It’s not even terribly good at remembering. What it is very good at is predicting.
Robert, I just had to move that “close blockquote” tag to the right place.
The best definition of what we are I’ve encountered is a device for predicting the future.
And the fact that symptoms generally kick in when an individual is way past reproductive activity means the gene(s) is/are not selected against.
I humbly suggest the process is pattern-matching. Learning means acquiring the useful patterns and thinking is matching them and reacting accordingly.
Eugenics is alive and well for certain devastating inherited conditions. The movie GATTACA suggested a future in which known detrimental genes are screened before conception. I believe the concept is that screening does not modify your progeny, but limits them to the best of what you are. Or something like that.
Not recommending this, but I would be surprised if it doesn’t happen.
We must disagree.
the memory is fantastic and I say thats all the brain is. A memory machine between our soul and our body.
Predicting is a very minor case in memorized data as I see it.
I’m surprised you dimish memory that much.
Why use words like pattern matching.
its just another for for memories dealing with memories.
The pattern was memorized. Simple. Then the next pattern was memorized and matched with the previous one. In all cases its a simple equation. Memory operation.
Why not?
I suspect you don’t like the idea of WHO is watching the memories or rather the soul immaterial.
Elizabeth,
Do you use GWAS data in your work? How about ENCODE data?
Sal