My mother claimed that she once taught my father a song well enough so that he could tell if a tune was that song or wasn't; he denied it. The purpose of this post is to report on a three generation project to breed musical ability back into my father's descendants.
My father married a musical wife. I married, in succession, two musical wives. My son by my first marriage married a musical wife. A few days ago, I observed the success of the project when my visiting grandchildren sang a song in tune—Frère Jacques as it happened. As the product of the first generation of the program I can recognize a tune although I cannot carry one.
Eugenics has a bad reputation because it is usually seen as the attempt of some people to control what children other people produce. I like to use "libertarian eugenics" to describe ways in which people control what children they themselves produce. Selecting one's mate in part on the basis of the children she will produce has been used to do it for a very long time.
That is, however, a primitive technology for the purpose; unless you are very lucky, you may have to sacrifice other desiderata to do it. Modern reproductive technology is beginning to provide better ways, ways of selecting, among the children you and your mate could produce, the ones you do produce. Carry the technology a little further, along lines suggested by Robert Heinlein in Beyond This Horizon, one of his less successful novels but one with several very interesting ideas in it, and we will be able to pick and choose among the parents' heritable characteristics, producing a child with my memory for poetry and my wife's musical ability but without her poor circulation or my genetics for a bad heart.
The results should be interesting.
79 comments:
Speaking of eugenics...the topic of selecting people for cold tolerance came up in a discussion on selecting orchids for cold tolerance.
The idea lends itself to an economics of genetics, i.e. how people exchange genetic materials for the purposes of creating new human beings, more than a libertarian eugenics. The genetic characteristics of a new human being are a function of the two mates plus random chance. Whether the two mates will organize to produce a new human being depends on their preferences. Sounds like economics to me.
Tell me if you think this is a flawed way of seeing the matter of eugenics:
As it is, any time you select your own mate and have even the vaguest intentions to mix your genes in sexual reproduction, you are participating in a probabilistic kind of eugenics.
Emergent technology allows us to greatly narrow the probability that our offspring will have certain traits, and in a Heinleinian future perhaps the probability of them having any available trait will be narrowable to 1, but the fundamental eugenic nature of reproduction remains the same. So, it's a matter of scope.
It's true that some of the bad reputation of eugenics comes from anxiety about some authority potentially dictating to others what traits are desirable or even allowable. But I think a deeper part of people's objection to eugenics comes from the feeling that it's unfair to overly narrow down the probability of some result whose probability is naturally either low or unknown. "It isn't sporting," as the British might say.
On musicality: I have never met anyone as of yet who would be really unable to learn to sing in tune. I had a lot of problems with that myself but then I met a good teacher and it just clicked. I'm nowhere near perfect yet, but I am improving...On the other hand I know at least two people whom I tried to teach rythm (both were at least not tone-deaf and one could sing quite well) and tt turned out to be surprisingly difficult. What I considered intuitive and extremely simple, they had trouble recognizing, they simply did not hear the difference between, say, an eight-note followed by a eight-note pause followed by a sixteenth note and a triplet followed by a triplet-pause followed by another triplet. We spent quite a considerable time trying to do that (and even simpler things when this did not work out) with little to no success. Then again, maybe I am just a lousy teacher :) Someone who did not just simply hear the difference at first and had to learn it himself, might be a better teacher for such people. I've always wondered if it is possible to learn absolute pitch. If you manage to learn one note, the others should be easy (as you can simply use your one note as a reference...although that is not exactly the absolute pitch, it is close).
Power Child: Show me parents that would not want the have the best child (from their standpoint) they could potentionally have, because it is "not sport". I doubt you will be very succesful at that :)
I guess it could have quite weird results though. Not necessarily bad ones, but wierd such as certain hair or eye colours being "in" at certain times so that for instance in one generation of people you would have say 70% green eyed and brown curly haired individuals, then 60% of those born cca 10 years later would be all black straight haired, with a light skin tone and so on. Since different clothes fit people of different hair/eye/skin colour differently, this would also have large impacts on more conventional fashion. Other effects are harder to judge. The question is if you can actually find "genes for musicality" and such and also if there is a necessity of tradeoffs in these areas. If it is possible to simply pick the best of all what the parents are offering then the only effect would be much healthier, smarter, better looking people in all directions. If tradeoffs are necessary the effects could be more interesting (although less desireable).
I love the line of thinking. Like Tibor, I'm having trouble with the specific example.
David, can you not sing Happy Birthday? When you're at a birthday party, you just go "mumble muuuumble mumble TO YOU" and hope nobody notices?
Tibor, I've encountered many people who claim absolute pitch can *always* be learned. However, it's a hard claim to evaluate. It's a bit like proving that a program will never terminate. Well, it hasn't terminated YET, but you just wait!
When I am at a birthday party, I say the words softly enough so that the fact that they are not in tune is not obvious. My guess is that there is a non-random correlation between what I am doing and what I am supposed to be doing, but not a very close one.
and we will be able to pick and choose among the parents' heritable characteristics, producing a child with my memory for poetry and my wife's musical ability but without her poor circulation or my genetics for a bad heart.
Except that the correspondence between genotype and phenotype, certainly for the first two characteristics, is virtually unknown. Nobody knows which genes (gene products really) are sufficient and necessary for poetry memory or musical ability, or even if genes are the most important factor in these characteristics, versus, say, environment.
And before anybody jumps to point out this or that association study (GWAS), the results of such studies provide no information about molecular mechanism and thus would be useless for guiding deliberative eugenics.
The limitation of doing what you imagine is not really technological, it's biological -- we simply don't know nearly enough about how most genes work, how they work together in a system, and what role environment plays in producing most phenotypes.
RKN: We don't have to know the mechanism if we can observe the correlation.
I'm a customer of 23andme. They routinely provide me information of the form "because you have such and such a genetic variant, your odds of such and such an outcome are higher/lower/the same as the average." In one case, they actually identified a higher probability of a medical problem that I had been recently diagnosed with.
There is no reason the same method couldn't be used for things such as musical ability. 23andme gives me the opportunity to answer questionaires on lots of things. With enough such data they can find out whether there are particular genetic variants that correlate with musical ability or other things. As complete genetic testing gets cheaper and cheaper and more and more data of that sort accumulates, any significant feature that does correlate with genotype should eventually be identified.
@Tibor Mach:
Right, it's one of those objections that's usually based on assumptions about what other people will do, never applied inward. I'm not sure I've ever heard the objection articulated exactly the way I put it above, it's just I get the sense that this dread of unsportingness underlies it. It's been a while since I watched it so I don't remember clearly, but isn't that kind of what underpins the drama in "Gattaca"?
You hint at another objection I have heard articulated, which is that as certain traits come in and out of style we'll get skewed populations with weird combinations of traits. (This is of course basically the same way evolution works except that "style" is dictated by procreative ability in the context of a given environment. By the time humans came along, bipedalism was mostly relegated to birds! Who'd have thought it would benefit a member of the ape family?! And that this ape would become mostly hairless! The combination is almost platypusian in its strangeness.)
Musicality clearly has a large genetic component, and interestingly is not tied to IQ. I also notice a divide between the ability to play an instrument well and the ability to precisely mimic recently heard rhythms/melodies. There's also a divide between general musicality and the ability to improvise creatively (i.e. not just patch together well-rehearsed "licks," but really speak articulately off the cuff through an instrument).
RKN: We don't have to know the mechanism if we can observe the correlation.
No sir. Not if your goal is to produce children with specific characteristics with a reasonable level of certainty.
If you want a child with exceptional musical ability, you need to understand, in some detail at a systems level, the molecular biology that produces that characteristic. Likewise, if you want to clone a high performance (HP) car, it's not helpful to note that a significant number of HP cars are red.
Without mechanism you don't know which genes need to be changed -- and which ones shouldn't! -- in the embryo's genome to achieve that characteristic. And that's assuming the characteristic is 100% genetically determined -- very many, of course, are not. Short of that, you're going to have a lot of angry would-be parents demanding their money back.
23andme is an association based analysis, as you seem to understand, the very kind I mentioned is not up to the task of informing gene therapy treatments. And it won't be up to the task with even more and more data. Which is not to say 23's analysis is meaningless, it's not, as I told a friend recently who asked me to review her 23 report -- she has a rare pulmonary disease. But 23's results will never provide the kind of information necessary to do predictive eugenics of the kind I think you imagine.
RKN:
I do not see why you need mechanisms. What you need to know is which genes code for what characteristics, not how they do it. Enough associational data can tell you that.
How much is enough is an empirical question. If the relationships are complicated enough, it might take a lot of data. But I cannot see any logical reason why it can't work.
Nor do you have to have "a reasonable level of certainty." It becomes doable as soon as you know that a particular genetic variant increases the probability of a particular characteristic.
RKN: I think you are refering to false correlations (good cars are red). These may of course happen and some parents may have the get the equivalent of red cars in their children...but then it will be noticeable from that new data, that the correlation was false, that the "red gene" does not seem to improve the musical talent for instance. As long as at least some of the correlations are correct, they will sooner or later be selected out. Of course, as you and David pointed out, a gene that would improve the ability by 0.0001% is probably not going to be noticed. But as long as there is a significant effect, you will eventually find out which genes are beneficial.
In fact, say that you are a mad scientist who knows nothing at all about genetics, only is somehow able to splice the aleles in any way. Then by simple trial and error - and bayesian inference - you would eventually see what is good and what is not. Without any understanding of why that is so (which is interesting and probably useful to know, but not necessary for this purpose). Fortunatelly, we do have some clues...Clues that can help avoid very nasty mistakes (which would prevent most parents from participating in the experiment). Clues gained both by previous inference of the current population and by biological analysis of the genes themselves, so we have a prior distribution of the good traits that is not entirely flat.
My diploma thesis was incidentally exactly about this problem. The application was on a stock market where you have the stock price modeled as a random process that depends on a parameter...which you don't know. If I skip the maths (even though that is the most interesting part :) ) you simply solve it by dividing your initial wealth (the population) among investors who each thinks that the parameter is some given value (parents, who think that the good gene is some particular one). You let then invest, observe the results through time and readjust the distribution of your wealth among them based on their performance (parents see how well the next generation does and bet on particular genes in the next generations based on that information). Eventually, you will get close to the real value of the parameter, or the right gene. I assumed that the parent are willing to participate in such a gamble...that may not be true if the process is too costly compared to its mean benefit. If it has a potential of improvement (whatever that means to those particular parents) of 5% and a potential of making things worse of 80% at a cost of 10 years of your wage, you might not be willing to do that. But that problem can be fixed by inventing a cheaper way to manipulate the genes and the odds can be improved at first by experimenting on, say, mice...which might not be very nice to the mice, but they don't really get to choose and if it can help humans, I think it is worth it. And these two things are exactly what is being done right now (except noone can actually manipulate the genes this freely...yet).
I have not seen Gattaca. But I know that if a Genie appeard and offered me an option to get the best children I could with the genes that I have, I would not think about it for very long and I would not care about being "sport" for a second. I think the same applies to most other people.
The second thing - That is what I said by possibly less desireable effects if there are tradeoffs. If you can simply improve the chances of any good trait, then it is no problem. Of course, if not, you might get a generation of perfect pitch musicians who however have few other skills, or something like that. That could be a problem.
As far as traits that are objectively neutral (neither harmful, nor beneficial) such as hair colour, height and so on, it could happen and would be quite strange to see generations differ so sharply as I described above. But I don't really see it as a problem, just something that we are not used to.
Musicality: What exactly do you mean by the ability to play an instrument then? Technique? I guess that is true, I just would not call anyone who has perfect technique but can neither reproduce songs he hears, nor improvise, to be able to play the instrument :) There most certainly is a difference between the rythmical and melodical abilities. Technique is of course yet something else, but a "musician" who just play notes that are given to him is like a "mathematician" who can solve a particular problem only after someone else has shown him the precise way to do that. That only shows he has some level of memory and that he can write and read (and the instrument player can read notes and knows the technique of playing his particular instrument).
David: I don't know if you want to improve your singing, but if you do, this helped me a lot: I sang a note that was false. The teacher played it (the false note) on the piano and then again played the note I was supposed to sing. I could hear the interval in between the notes, I knew my margin of error and that helped me considerably to get the right note, even at the beginning I usually got it after one false try. After some time, I simply got the right note on the first try...but only at a certain pitch at first. It seems you develop a memory for notes even if you don't have a perfect pitch (which I certainly don't) so that in one octave I got the notes right each time, in a different octave it was more difficult to sing the same notes, because I had not practiced that octave (or not necessarily octave, some other "critical" interval, or even single notes as well). Once I was able to do that, I practiced (and still do) singing with a guitar. I simply sing the melody and play it simultaneously on a guitar. Then I can better hear when I am out of tune...Once I can do this with ease, I play the chords in the song instead and try to keep the melody right. Sometimes I record myself to check if it actually is right. Some people probably learn it more easily, but more you do it, the easier it gets next time. Of course, if there is no incentive for you to improve, you won't even with a great talent. Perhaps your mother, if she actually wanted to teach your father to recognize songs, should have awarded him with a chocolate cake (or something liked) each time he got the song right :)
Tibor,
You're assuming a way, way higher level of talent than actually exists. Dad could probably train himself to sing if he really wanted to - but he'd have to use a tool like Rock Band that tells you when you're on, because, if he's anything like as bad as I was when I started learning to sing (and I suspect he's worse), he probably can't hear whether he's matching in the first place (I couldn't), let alone how far he's off by. Hearing what the interval is is advanced enough I still can't necessarily name you a sixth if I hear it. (Fifth I might be able to - my instrument is the harp, and the harp uses fifths a lot. Third I couldn't. A minor third I might still hear as just a discord - I used to do that quite frequently.
And I'm the second generation of this breeding project.)
Note also that being able to reproduce a song accurately with one's voice is not the same thing as being able to hear it in one's head, know it when one hears it out loud, recognize minute variations, have firm opinions about how it should (or should not) be sung, and so on. The statement, and-I-quote,
"I just would not call anyone who has perfect technique but can neither reproduce songs he hears, nor improvise, to be able to play the instrument :)"
- excludes a huge number of quite serious musicians, who despite being quite good, able to play with emotion, etc., play nothing or almost nothing without the sheet music in front of them - let alone reproducing songs they hear, they don't memorize their own music. Yet they're perfectly valid musicians. It's the difference between classical training and folk training; a classical musician can sightread and may or may not be able to play from memory (either of things he's played, or things he's only heard), while a folk musician memorizes, can probably play anything s/he knows to sing, but often can't sightread to save his/her life. And improvisation is a whole different skill, that on one level a 3-year-old can do, and on another - depending on your genre and instrument - may take quite a lot of independent work and study if you want to do it well, unless you naturally have that talent, which many people don't.
... Sorry. Just... I envy your talents. Please don't assume everyone has them. A lot of people don't. It's just easy not to notice, because people as bad at music as me usually don't even try - if I didn't have a musical mother, I probably would have given up long since.
i don't select my lovers by how good they are at playing the piano, no one does and no one will.
breeding children in the way you propose is extremely undesirable and MUST be stopped before it is too late. what if someone decides to breed serial killers, or breed ruthless soldiers, what if people begin to judge each other based on the results of genetic testing, and those who fail the test are forever guaranteed unemployment?
it is ironic that as a jew yourself, practising or not, you are proposing allowing people to breed a perfect, master race.
if you want to quote conjectures about the possibilities of genetic engineering then look towards better science fiction than heinlein, look towards star trek. if it is one thing that star trek has taught us is that genetic engineering will lead us to no good.
^lol
Rebecca: I think you're right about reading sheet music. I can do that a bit, but unless the piece is a well...a piece of cake, I cannot do that in real time. Improvisation is a bit hazy. It is like that famous porn quote "I cannot define good improvisation but I know it when I hear it." But like other things, it can be learned...by doing it a lot. Which is, sadly, what they often don't do much at the classical music conservatories. I guess, you don't need to be able to improvise if you play in an orchestra. But as soon as you try to make your own pieces or just arranges, it comes in handy, because it helps with the composition significantly. I may be a bit biased as I have mostly learned to play in what you describe as the "folk way" (even though I can read the notes a bit and had teachers). Someone else might object that I am not a real musician since if they dropped me in an orchestra and there was a sheet of music in front of me I had never seen before, I would have a very hard time to catch up with the band. But improvisation is the main creative element and for me, it is one of the main things that make playing music fun for me. It is not the only creative part. Even if you play sheet music (or a memorized part) you can play it in various ways, all of which are correct, but all of which have a different "feel". When I think about it, this may be one other things humans are better at than computers. Computers can play right (or you can even program slight mistakes), but they cannot play with "passion", "feeling", "groove" or whatever you call that...mainly because humans are unable to describe it precisely themselves :) What I meant when I said I would not call someone who only has good technique and nothing else a musician, was a person who is not only unable to improvise or repeat what he hears (it is really hard to do that with someone like Jeff Beck who just makes the sounds in a very original way and even if he plays something simple, it is just hard to figure out how he did that...or when the piece is simply very complicated and fast) but cannot play with "passion". Then that person is no better than the computer...worse, really, as the computer simply will not make any mistakes at all and offers its services really cheap. Also, computers can really repeat what has been played (exactly, since they can record it) and maybe there are even programs that can teach them limited improvisation that does not sound all that awful. But they don't have the "groove", so that is the main difference between a bad and a good musician, I guess.
Rebecca:
Also, a great tool to learn to hear intervals:
http://www.musictheory.net/exercises/ear-interval
there are other useful tools at the website as well (chord recognition for example). A great way to learn the intervals is to assign a song you know to each one of them. So, for example the perfect fourth is the first two notes of Oh, Tanenbaum (Also, I think Auld Lang Syne, but I can't remember the melody of that right now). It takes some time, but it can be learned by anyone, I think. I can now recognize most of them when played up, almost none when played down, I have not practiced that yet. The people I play with are far better than me at this (one is a son of a guy who has perfect pitch and he also is able memorize at least some notes perfectly)...that may be why I play the drumset :) (apart from liking the instrument). It is hard to tell, but I think I am not exceptional at this in any way. I have naturally good rythm I think, but I have to really train my ears hard for pitch and intervals.
You also don't need to recognize the precise interval between the false note you sing at first and the right note you are supposed to sing. I think you just need to know if it is higher or lower. And even that you don't have to hear precisely at first. Maybe this method won't work for someone, but I know I tried to learn to sing in tune for a while (by myself) and then I went to the teacher who tried this method (and told me it helped quite a lot of other people who seemed to be entirely tone-deaf at first) and it suddenly worked out relatively fast.
I do not see why you need mechanisms.
Because if you wanted to build a person with specific traits, you would need to know how the biological traits work at a systems level.
What you need to know is which genes code for what characteristics, not how they do it.
Oh that biology was that easy!
This is precisely the problem, for the vast number of human traits, certainly the kind of traits you were talking about, we don't know which genes code for which traits. Hell, even for a simple physical trait like eye color we don't know all the genes involved. And you don't get which genes code for which traits from a 23 analysis, or what environmental factors are important. What you get is what SNPs associate with a particular trait across a limited cohort of samples. And very many of the significant SNPs 23 detects aren't even in the protein coding region of a gene, in many cases they are thousands of kilobases away.
Enough associational data can tell you that.
No, it most certainly cannot.
Of course, as you and David pointed out, a gene that would improve the ability by 0.0001% is probably not going to be noticed.
The problem with GWAS analyses ability to detect rare variants (versus common) is well documented in the literature, although I expect this will be ironed out before too long.
But as long as there is a significant effect, you will eventually find out which genes are beneficial.
Except when the variants don't appear in a gene, as many don't. Then what?
There are answers to this question which are outside the scope of a blog comment, and would require (at least) some understanding of the details of the central dogma of biology, i.e. DNA-->mRNA-->protein. But in any case, they wouldn't overcome the inherent limitation of information coming from GWAS, and would underscore the importance of what I've been saying about biological function.
I can assure you that it would be wildly misleading, to say the least, if anybody said to you that by merely assembling a large enough "parts list" of DNA variants we will eventually understand the biological basis of trait X, and -- and! -- eventually with that information alone be able to engineer humans who express trait X. [1]
That may play well at a TED talk, but not among biologists who know better.
1. A similar claim was made to justify the human genome project and how it would lead -- finally! -- to us conquering many important human diseases. Bzzt.
Anonymous:
As you may or may not know, a real world example of selective breeding to eliminate an undesirable trait in the phenotype is provided by Ashkenazi Jews. Taysachs disease is a serious problem due to a recessive gene. There has been a long term private, voluntary project, to permit couples who are considering dating to find out if they are both carriers, in order not to start a relationship that might lead to children exhibiting the result of having both recessives.
So I think your idea that being Jewish is a reason to be opposed to voluntary selective breeding is bizarre.
So far as your other points, one may not choose one's lovers by their musical ability--I didn't. But one might well choose the wife one plans to have children with by whether her characteristics are ones you would like to see in your children--along with other desiderata. And once our technology is good enough so that we can select which of our traits go to our children, we can do that after mate selection--selective mating is the primitive version of the technology.
As to the rest of your objections, they strike me as paranoid. I think it much more likely that parents would choose to avoid giving their children the traits that produce ruthless soldiers than deliberately give them those traits, so the sort of libertarian eugenics I have described makes the outcomes that you describe less likely, not more. You seem to be confusing eugenics done by parents for their own children with eugenics imposed by some people on other people's children. I think I made it clear in the post that it was the former I was discussing.
RKN:
"I can assure you that it would be wildly misleading, to say the least, if anybody said to you that by merely assembling a large enough "parts list" of DNA variants we will eventually understand the biological basis of trait X, and -- and! -- eventually with that information alone be able to engineer humans who express trait X."
I wouldn't claim that one would understand the biological basis of trait X, but I would claim that, with enough data, one could engineer humans more likely to express trait X.
If your account is correct, how is it that human being have succeeded in selective breeding of lots of other species for specific characteristics? The information "this individual has characteristic X" is less informative than "these individuals who all share a particular gene variant have characteristic X," not more informative. Yet with only the former information people have been doing successful selective breeding of plants and animals for thousands of years.
RKN: Your prior distribution for the variants that don't appear in the gene is flat at first. I guess (although I am not a biologists or a genecists...even though I will probably will be working with some over the period of the next few moths and years) that you can get at least some information about them by biological and chemical analysis. I'm not saying that you can discover that "this is a gene that will do so and so" but that "this is a gene that will probably affect this part of the body and the effect is more likely to be positive". That gives you a better than flat prior distribution. Experiments on animals can improve it even better. When you exhaust these options you will hopefully end up with good enough estimates so that people will be willing to take the bet...and then you collect data and make posterior distributions after each generation. In the meantime, you also include new discoveries in the biological aspects that improve it even better. But as long as there are some genes that meet these criteria, I really don't see anything that would make it impossible. Again - I did not claim that will help you understand the biology behind it. It won't. But you don't need it. I did claim that it will help you get the right result...
it is like adaptive AI works. The computer cannot think, it does not know why this or that lead to the desireable result. All the program it is running needs to know is when it reached that result...and since it memorizes its previous actions, it will make those actions that lead to the desireable result more likely to take place in the next game...eventually, it will find a strategy close to the perfect one...without any understanding of why that strategy is so good. Or without any understanding of anything at all, really.
i am far too young to have children or a wife but i hope when i do come to these matters i will not be marrying a woman because she is good breeding stock. i will give you that intuitively/subconsciously attractions probably are partly based on judging the healthiness of the mate.
by merely suggesting what you are suggesting you inevitably come to a full scale master race breeding program. at first parents simply screen their embryos for defects, after a while they screen for a huge list of problems, after a while this becomes common, it is then thought that people who do not do such things are hurting their future children, a law is passed to make screening compulsory. after a while people star wondering why not actively introduce positive traits to their children, the technology is there already, so they do so. of course healthcare is entirely government run. it becomes a political question of what can be enhanced. eventually government mandates exact program of selection and enhancements. are we not at a master racing breeding program now? do not forget that america was the country from which the nazis took their ideas about eugenics from: three generations of imbeciles is enough Buck v. Bell.
you may denigrate me by calling me paranoid but that still does not take away from the full horror which your idea of voluntary eugenics necessary lead to.
no, all of this kind of genetic engineering technology must be banned while we still can do it, before it is too late. don't become a real life Arik Soong.
@Anonymous (the paranoid one):
I'm pretty agnostic about the long-term benefits of eugenics, but your slippery slope argument against it is worse than useless.
Slippery slopes are all around us. Yet, we perch at the top of most of them, not sliding down. This is because there is a great deal of consensus among human beings about reasonable limits and moderation. I can give you numerous examples if you're interested--just ask.
We've been selecting our mates for desirable heritable traits (among other things) for unknown millenia. There have been numerous popular eugenics movements in that time, many of them much more popular than any such movement today, yet none of them led to centrally-authorized selective "master race" breeding programs. Even Hitler's attempts at this were poorly articulated and never realized to any large degree.
I wouldn't claim that one would understand the biological basis of trait X, but I would claim that, with enough data, one could engineer humans more likely to express trait X.
Unclear what you mean by "enough data." Sure, if I had all the data, if I know all 3 billion base pairs of a musical virtuoso (something already technologically feasible -- full genome scan), then sure, function be damned, I just clone me a virtuoso once the problems with human cloning have been worked out. Although even there I would expect wide disagreement on just how "likely" it would be that the clone expresses musical virtuosity, given what we know about the importance of environment (nurture), especially for complex traits. But GWAS analyses don't involve full genome scans or provide the kind of data needed -- even if there was more it -- to engineer preferred traits a la carte, and at the same time avoid incidental, deleterious traits. To do that, one would need a much better understanding than we presently have of gene (product) function.
If your account is correct, how is it that human being have succeeded in selective breeding of lots of other species for specific characteristics?
The kind of future technology you imagined where parents would be able to "pick 'n choose characteristics" would not be based on principles of Mendelian inheritance, principles discovered over 150 years ago long before there was any agreement on what a gene even was, far less GWAS, so I don't understand how this is in any way at odds with "my account."
@Tibor Mach
My point isn't that there is no biological relevance to GWAS results, there is, hypothesis generation for instance, to guide certain validation experiments in relevant cell or animal models. Without independent validation most of my papers wouldn't have passed peer review, and the ones I review likely wouldn't have gotten past me without it.
And just a word of unsolicited advice to you should you find yourself working with biologists -- take it from me, someone with a PhD in this area who now works to discover biomarkers of complex human disease: When working with biologists, especially as a computer/math guy which it sounds to me like you are, never let a biologist here you say "understanding the biology is unimportant". If nothing else, a major CLM.
RKN: Cloning the virtuoso doesn't give the desired result--parents want a child of their own. But by the time we have the sort of technology I described, a full genome scan should be cheap, so our statistics can be on millions of them, for each of which we would have a complete genome and a list of characteristics we were interested in.
If a relationship is sufficiently complicated that won't identify it, but it should be sufficient to spot any correlation with a single gene variant and many that depend on more than one.
What am I missing?
My point about selective breeding was that it was done with no knowledge of the underlying biological mechanism of the traits they were breeding for--just the knowledge that the offspring of individual X was more likely to have a characteristic X had than a random individual. Yet it worked. Statistics on lots of complete genomes should let you do much better than that.
RKN: I'm sorry if you took that personally, I did not mean to say that biology is unimportant (by the way, I am currently studying coalescent processes which have a lot of applications in genetics and I will probably be working with some genetists...actually I already have one seminar together with some of them, although it is really about the mathematical models and not the applications). Of course it is important! But biological knowledge of what this and that actually does is not a necessary condition for you to reach an improvement. It is sufficient, but not necessary. Take a look at the AI learning analogy again. Some smart game theorist can prove that a certain strategy in that game is optimal. That is a great result. He could then simply program the computer to follow that strategy. But there is another way, that leads to very similar results, albeit not immediately, and which does not require the AI (or anyone else) to know anything about which strategy is good. The almost-optimal stategy is then reached simply by trial and error ... and subsequent readjustment of probabilities of each turn based on whether this and that move eventually ended up in winning or losing the game. Essentially, by bayesian inference. Now, of course the way there can be very long. If the game is very complicated and each move has only a tiny chance of changing the outcome, you may not be able to reach something meaningful in a reasonable time without some prior knowledge...which, in the case of a game can be provided by partial mathematical results about good strategies, in the case of genetics by partial knowledge of the inner workings of the genes and their biochemical processes.
Of course, if you know exactly what the genes do, then you don't need the trial and error method. But since you yourself have said that is (at least today and in the near future) extremely hard or even impossible, then the bayesian approach is a good substitution of the missing knowledge. And it strikes both ways - if you observe that something leads to better results with a good deal of probability, you focus your biological research efforts on that...and hopefully reach the analytical results sooner than you would have otherwise.
Oh I see, parents want their "own" children, except you imagine a future where tweaks (one or many) are made to certain nucleotides in the zygote's genome, changes that statistically significantly associate with a certain trait, say exceptional musicality, and you believe this genetic engineering approach alone would produce a child more likely to express musicality, without introducing other deleterious biological effects.
Does that fairly describe the "pick 'n choose" reproductive technology you had in mind?
If it does, then what you're missing, still, is the fact of poor correspondence between genotype and phenotype, especially but not only in complex traits.
"Does that fairly describe the "pick 'n choose" reproductive technology you had in mind? "
No.
As should have been obvious from the original post, where I wrote:
"ways of selecting, among the children you and your mate could produce, the ones you do produce."
and, later,
"producing a child with my memory for poetry and my wife's musical ability but without her poor circulation or my genetics for a bad heart."
The method you describe would produce children with characteristics we wanted but neither of us had.
Then what was this supposed to mean?
Carry the technology a little further, along lines suggested by Robert Heinlein in Beyond This Horizon, one of his less successful novels but one with several very interesting ideas in it, and we will be able to pick and choose among the parents' heritable characteristics
What do you mean by "able to pick 'n choose parents' heritable characteristics"? What future technology? Please be specific.
The method you describe would produce children with characteristics we wanted but neither of us had.
Uh, actually, it wouldn't, certainly not with anywhere near the likelihood you imagine, or without introducing deleterious effects, for reasons I have made clear.
Heinlein's technology in the novel let the parents determine which genes were in each of numerous sperm and each of numerous eggs, using an ingenious trick to avoid the problem that you can't analyze the egg or sperm without destroying it. They could then choose which sperm and which egg to combine and implant.
So if you know that a particular chunk of the husband's genome correlates with a characteristic you want, you eliminate sperm that don't contain that chunk, then repeat for the next desired characteristic until one sperm and one egg are left.
"Uh, actually, it wouldn't, certainly not with anywhere near the likelihood you imagine"
Given that I had just explained that the method you described was not what I was proposing, I cannot see how you formed an opinion of what likelihood of its producing results I was imagining.
As you surely know, a primitive version of what I am describing already is in use--producing multiple fertilized eggs, letting each get to the 8 cell level, at which point a cell can be pulled out and analyzed without injuring the rest, analyzing it, and selecting one of the eggs to implant that does not have whatever serious genetic defect is being avoided. What I am proposing is to do the equivalent one step earlier, for sperm and egg separately, with enough more knowledge to not only avoid very serious negative consequences but select in favor of desired characteristics.
While I never heard your mother speak, I have noted that your father's grammar was excellent. Yours, on the other hand, is sorely lacking for a grad of U of C and a professor at a law school.
A physicist myself, I have seldom met or listened to another "native speaker" physicist who seemed not to have mastered English grammar, except perhaps for Feynman.
The question is: how could it be that your musical handicap is inherited but your language skills not?
Heinlein's technology in the novel let the parents determine which genes were in each of numerous sperm and each of numerous eggs, [...]
So if you know that a particular chunk of the husband's genome correlates with a characteristic you want, you eliminate sperm that don't contain that chunk, then repeat for the next desired characteristic until one sperm and one egg are left.
I'm afraid you've lost me. Naturally, of course, the haploid DNA inside every sperm cell of a would-be dad is the same. Same for mom's DNA in the eggs. So what technology do you imagine is going to create the variation across the sperm/egg cell's DNA you'd need as the basis for this selection? And how would this "annealed" genome be any more the father's "own" than the one resulting from the genome editing method I described, the one you said didn't describe your approach?
To say nothing of the fact that it's not "chunks of DNA" that associate with traits, not at least as are detected by modern GWAS, but individual nucleotides, usually a few to a few hundred or less. So even by this piecemeal construction approach of yours, which I still don't understand, you're going to end up with very many "chunks of DNA" making it into your final genome for which there is no known association to any particular trait. You can't just smash any old "chunks" of DNA together and make any kind of chromosomes you want and expect the gamete will be competent in fertilization. (Please see homologous recombination).
You appear to think: DNA-->Phenotype. It's really not that simple. Not even close. I don't what else to tell you.
Given that I had just explained that the method you described was not what I was proposing, I cannot see how you formed an opinion of what likelihood of its producing results I was imagining.
Fair enuf, I retract the portion: "likelihood you imagine."
Suffice it to say I strenuously disagree with you; the method I described, the one you just now said did not describe your approach but would produce children with the characteristics we wanted, I say would in fact not likely produce children with the characteristics we wanted. How's that.
As you surely know, a primitive version of what I am describing already is in use--producing multiple fertilized eggs, letting each get to the 8 cell level, at which point a cell can be pulled out and analyzed without injuring the rest, analyzing it, and selecting one of the eggs to implant that does not have whatever serious genetic defect is being avoided.
Here you're describing a post-fertilization selection process, after recombination and crossover have occurred, which is stochastic in terms of outcome, and used to screen for allele patterns that confer known diseases. That is very, very different from the kind of forward engineering approach you described, which as I said I still really don't understand, and I'm left wondering if you do either.
RKN:
Naturally, of course, the haploid DNA inside every sperm cell of a would-be dad is the same. Same for mom's DNA in the eggs. So what technology do you imagine is going to create the variation across the sperm/egg cell's DNA you'd need as the basis for this selection?
This is not true. If it were, all siblings would be clones.
If I've understood how David describes Heinlein's technology (correct me if I'm wrong), is that since all of the father's genetic material exists in the diploid primary spermatocyte, and none of it ideally is lost or duplicated during spermatogenesis, you can determine what genes the 4th sperm cell carries by sequencing the genomes of the other three and 'subtracting' them from the original diploid genotype.
Super clever. Solves the problem of genotyping a cell without destroying it.
Also, David and Tibor's inferential method of eugenics is simply a step up from natural selection. Nature can select traits without identifying any genes or knowing anything about gene interactions; surely humans, armed with data and tools for selecting sperms/eggs, can make incremental improvements as well.
This is not true. If it were, all siblings would be clones.
Nonsense. The reason siblings aren't clones is due to the stochastic nature of recombination and crossover.
"I'm afraid you've lost me. Naturally, of course, the haploid DNA inside every sperm cell of a would-be dad is the same. Same for mom's DNA in the eggs. "
Nonsense. Each sperm cell contains half the DNA of a full cell, as does each egg cell (aside from the mitochondrial DNA).
Joseph Miller has correctly guessed Heinlein's elegant solution to the problem of figuring out which genes are in a given sperm or egg without damaging it.
RKN: You can find a fairly detailed explanation of meiosis, the process by which sperm cells and egg cells are created, here:
http://faculty.clintoncc.suny.edu/faculty/michael.gregory/files/Bio%20101/Bio%20101%20Lectures/Meiosis/meiosis.htm
Note, in particular:
"Meiosis functions to reduce the number of chromosomes to one half. Each daughter cell that is produced will have one half as many chromosomes as the parent cell.
Meiosis is part of the sexual process because gametes (sperm, eggs) have one half the chromosomes as diploid (2N) individuals"
Nonsense. Each sperm cell contains half the DNA of a full cell, as does each egg cell (aside from the mitochondrial DNA).
Yes, half the DNA, one set of chromosomes (not two), which is what haploid means, which is what I said. Did you not bother to look it up?
Now, as it turns out, I think I owe Mr. Miller an apology for replying too hastily and brusquely. In fact, the four spermatids, i.e the four daughter cells resulting from two stages of meiosis, are not in fact isogenic as I previously said. When I wrote that I had in mind the male germ cell, which, as he said, is diploid. However, I was correct insofar as the genetic diversity among the spermatids is due to recombination & crossover.
The remainder of my criticisms stand.
RKN: You can find a fairly detailed explanation of meiosis, the process by which sperm cells and egg cells are created, here:
Srsly?!!
Nothing wrong with your sense of humor I see.
Considering that you wrote:
"Naturally, of course, the haploid DNA inside every sperm cell of a would-be dad is the same. Same for mom's DNA in the eggs."
I thought pointing you at an explanation of the relevant biology was appropriate. Your statement was made about sperm cells and egg cells, not about germ cells. And it was the sperm and egg that were relevant to the exchange.
Quite aside from replying too hastily to Mr Miller, do you agree your reply to me was wrong? The male sperm cells are not identical, hence the procedure Heinlein proposed is, given enough information and an adequate technology, workable. Hence your response made no sense.
you should watch the film Gattaca, the results of voluntary breeding like that are not good. http://youtu.be/azpx_EsreGE
"I belonged to a new underclass, no longer determined by social status or the color of your skin. No, we now have discrimination down to a science."
"They used to say that a child conceived in love has a greater chance of happiness. They don't say that any more."
I *said* the DNA in a sperm cell is haploid, which is true when sperm cell refers to spermatid. However, I also said all four of these sperm cells are isogenic; that part was not true. I had the germ cell in mind when I wrote the latter, and referred to it (mistakenly) as a sperm cell, which shouldn't be *too* unforgivable given the myriad cell types (6) involved in spermatogenesis generally, one might understandably use "sperm cell" to refer to any one of the intermediate cell types.
Now, let's dissect your response to that. Your reply pertained to the first part of what I wrote, to wit: "Nonsense. Each sperm cell contains half the DNA of a full cell." Which of course is what haploid means, but I don't think you understood that, you pretty clearly intended it to disabuse me, a motive confirmed in a followup comment where you provided, evidently for my edification, the meiosis link.
Quite aside from replying too hastily to Mr Miller, do you agree your reply to me was wrong? The male sperm cells are not identical, hence the procedure Heinlein proposed is, given enough information and an adequate technology, workable. Hence your response made no sense.
I've said twice now, here and in a prior comment, which part of my reply was wrong. But the reply you're referring to was multi-part, most parts remain relevant in spite of the mistake and contain questions that have largely gone unanswered by you, e.g., "chunks of DNA" associating with phenotype rather than what GWAS-type data really provide, etc.
In any case, it remains unclear to me how the technology could serially isolate a single spermatid with the precise set of nucloetide variations that associate with a trait, nevermind that even if you could it's unlikely to produce the kind of child you want, for other reasons I've stated, or why, when it comes to performing the same in eggs, you think you wouldn't run out reproductively competent eggs long before serial selection isolates the precise maternal DNA you think you need. Women are born with a fixed number of eggs, only a few of which mature during ovulation to the stage you'd need for the selection process. In fact, it's worse than that, a quick review of oogenesis indicates that phase II meiotic division leading to the competent oocyte, the haploid equivalent of the spermatid, doesn't complete until after fertilization, but your process would need to screen before fertilization.
No, at the level of genetics the approach is dubious at best for producing the goal you want; at the level of technology it reduces to arm waving -- someday technology will work it all out.
RKN said to D Friedman:
Here you're describing a post-fertilization selection process, after recombination and crossover have occurred, which is stochastic in terms of outcome, and used to screen for allele patterns that confer known diseases. That is very, very different from the kind of forward engineering approach you described
I think this was a misunderstanding and may now be cleared up. The libertarian eugenics that David and Tibor were talking about never required any kind of "forward engineering"; it only required knowing that people with gene X are more likely to have trait X, and selection of sperm/eggs with gene X.
This is exactly how natural selection would operate if all other alleles were lethal or caused infertility.
Paranoid Anon:
I wonder if you're consistent enough to also oppose mate selection even when it's not deliberate? Should basketball players be forbidden from having children with tall women? It seems clear that a precedent of a ban on genetic engineering in humans would more likely lead to a dystopia than allowing parents to do so.
Who, exactly, loses if parents choose to select their mates/children on the basis of being informed about their genetic quality?
The libertarian eugenics that David and Tibor were talking about never required any kind of "forward engineering"; it only required knowing that people with gene X are more likely to have trait X, and selection of sperm/eggs with gene X.
Except that's precisely the "requirement" we don't have -- especially for complex phenotypes, say "musicality", we don't which genes code for the trait, you don't get it from GWAS because the SNPs frequently don't reside in or anywhere near genes, some aren't even located within conserved stretches of the genome, and getting lots more GWAS data won't overcome this for the well-known reason of a generally poor correspondence between genotype and phenotype [1]. So you don't have the information you need to guide the sperm/egg screening "technology", notwithstanding the practical problems of screening maternal DNA I alluded to in my prior post.
Look, if all this really boils down to is: Musical children will more likely be produced by one (or two) parents who are musical, okay, shrug, I don't have any important objection to that much. Although I'd be dubious until provably convinced otherwise that nurture doesn't have a great deal to do with it.
1. The notable exception being certain diseases conferred by a single mutation in a protein coding gene, e.g. cystic fibrosis. (And even there, there is considerable heterogeneity of phenotype observable across patients with the same genetic disease).
RKN: You scan the genetic information of a few million people. Note which one of them are musical (which you measure by how well they recognize notes, intervals, rythm variations and so on). Then you look for statistically significant differences in the genetical information of the musical people with respect to the whole population. Then perhaps you find out that a particular string of DNA happens to be more likely (say 20% more likely and you would need fewer than that) to be present in the DNA of the musical people. That is a correlation, not causation, sure. But while correlation is not a suficient condition for causation (unless you can be sure to have e.g. a normal distribution :) ), it is a necessary condition. So then (now suppose you can read the information in each sperm without destroying it as described above or perhaps in some other way) you look for the sperms that contain that particular string of DNA. You use one of those to fertilize the egg. Then you are by no means guaranteed that the child will be more musical. But you have improved the odds....it can of course turn out that the correlation was bogus, that there was a whole different reason for musical people to have that gene which had nothing to do with their musicality, but was in some unpercieved way related to that group of people. Well, then you know more for the next generation, you try again.
What I totally agree on with you is that there definitely can be traits complicated enough (and musicality could be one of them) that it is really hard to find anything different and statistically significant in the population of musical people...so you simply don't know where to start. That can also change in time as you accumulate more data, get better tests, find breakthroughs in biology and so on (maths improve as well...in fact, a lot of improvements in computing are not just by more brute force, but simply by clever people coming up with faster alghoritms...a great example of that is the difference between those old F-something US military aeroplanes from the 80s that were really edgy and the newer generation of "invisible bombers" which really have no edges at all. That is because the mathematical model is probably identical or very similar, but the approximation alghoritms improved so that now you get closer to the optimal shape, whereas you only had a very rought estimate back then).
So maybe something is really really hard. But I don't think that is a good reason to dismiss the whole idea. If you can improve 20% of the stuff and break even in the rest, it is still a pretty good deal (and would probably get better as the human knowledge improves and so the odds).
Your point about the scarcity of eggs:
Sure, you cannot choose among millions of them. In theory the women could simply freeze the eggs at every cycle, so they would get a few dozen to choose from once they decide to actually concieve a baby. Provided that the technology to do that is safe enough to do that each time (which I gather is not the case today, but maybe I am wrong) without too much of a risk and of course cheap enough as well. Even if it is not, even if you are stuck with one egg (or more realistically, a few), you can still at least pick the right sperm. Again, you don't need perfection, it is sufficient to improve the odds and if you can squeeze out (quite literary, really :) ) the best of one parent at least, you have done just so.
Also one thing that might be worth mentioning - it seems to me you understood David's description as "splice all the sperms, recombine then in the desired perfect sperm" whereas the idea was "select the one sperm that contains the most of what you want it to contain among the ones the prospective father has". The first one would not only be much more difficult, it would actually not satisfy the condition of having your own babies. But the second one does.
Joseph Miller:
Another point I think is worth to mention is that even if you ban something de jure, you are not going to do it de facto. Heroin is illegal everywhere, almost everywhere, you have serions penalties for its use or at least distribution, still it gets made and distributed.
Now having better children is something a great deal more attractive to many people than heroin is. If the relevant technology existed, you would have a black market on a huge scale with all the undesireable consequences of that. And instead of pretty much everyone having the possibility of getting the kinds they want, you would make it much harder and more expensive (also more risky, but that is another thing). That would only increase the division between the "enhanced breeding" and "regular breeding" people which I reckon is what most people who don't like the idea are afraid of (apart from the fashion thing).
So even if there were real issues of that technology, once it gets developed - and it will be developed if it can be for the exact same reasons the black markets arise despite the bans...people want it and are willing to pay for it - prohibition would really only make things worse.
The only real way to stop that (unless it turns out to be physically impossible, but I doubt that) is to persuade everyone that it really is a great danger and a big problem...given the nature of the thing, well, I don't think anyone could suceed in that. Perhaps if he starts his own anti-enhancing religion. But sacrificing "better" kids is still a much bigger deal than sacrificing, say, pork, beef, alcohol, coffee (LDS), blood transfusions (Jehovists) or a lot of money (Scientologists). So one would have to be really good at the prophet job. I guess even that old jewish carpenter would have a hard time with this one.
@Tibor Mach:
Some time ago I came up with a Rule of Prohibition: basically, that simply prohibiting something for which there is high demand is not enough to ensure a dangerous black market.
A lot of the existence of "informal markets" (like those built up around illegal drugs) depends not only on the level of demand for the product, but also with how easily the product can be concealed, transported, substituted, used discreetly, whether the product is transferable, and especially with what kind of culture exists among the people who demand that product.
So, while I tend to agree that prohibiting technologically-assisted gene selection would be a net negative on utility, I don't agree that it would necessarily have "all the undesirable consequences" of something like heroin prohibition.
Let us remember that there are some prohibitions on things which are a net positive, and do not contain all the negative consequences of the drug war.
RKN:
I mostly agree with you, but I think you were being a little too dismissive.
Look, if all this really boils down to is: Musical children will more likely be produced by one (or two) parents who are musical, okay, shrug, I don't have any important objection to that much. Although I'd be dubious until provably convinced otherwise that nurture doesn't have a great deal to do with it.
Well, I agree except with that part ;)
As a rule, I don't think evolution leaves much for nurture to interfere with.
Tibor:
That is, as usual, a good point.
That would only increase the division between the "enhanced breeding" and "regular breeding" people which I reckon is what most people who don't like the idea are afraid of (apart from the fashion thing).
In that spirit, I propose another mechanism for a Morlock-Eloi outcome of banning libertarian eugenics:
Smart, able people tend to prefer to mate with other similarly smart or able people, but in today's world can only use status-related extended phenotypes. This logic seems to result in increased inbreeding within social classes.
If all of your information is only tenuously related to genetic quality, it makes sense to prefer a mate with many weak signals, rather than search for a single, strong signal a la illegal genetic testing.
On the other hand, allowing libertarian eugenics should result in all humans, even poor or otherwise unfortunate ones, forever becoming higher in quality.
RKN: You ask about ""chunks of DNA" associating with phenotype rather than what GWAS-type data really provide, etc."
The point of my "chunks of DNA" was that the process doesn't randomly select genes--genes that are close to each other on the strand are likely to stay together. So you don't get to select single genes in Heinlein's method.
But you do get to select a chunk that contains a gene you want, and to some extent, among such chunks, one that doesn't contain genes you don't want or does, if you are lucky, contain other genes you want.
"In any case, it remains unclear to me how the technology could serially isolate a single spermatid with the precise set of nucloetide variations that associate with a trait"
What part is unclear? The "precise set"? That's making the best the enemy of the good. You can't do a perfect job of selecting for (say) musical ability, both because the parents may not have the relevant gene variants and because you can only scan a limited number of sperms and eggs--more limited, as you point out, in the case of eggs.
But you can do a whole lot better than random chance.
"I wrote the latter, and referred to it (mistakenly) as a sperm cell, which shouldn't be *too* unforgivable given the myriad cell types (6) involved in spermatogenesis generally, one might understandably use "sperm cell" to refer to any one of the intermediate cell types."
But not when responding to an argument that was explicitly about the sperm and egg, not about other cells. If you followed the description of the technology, your answer made no sense unless you really believed what you wrote--that all the husband's sperm were identical and all the wife's eggs. Presumably your answer is that you hadn't followed it.
@David - your thoughts on Fama, Hensen and Schiller?
Power Child:
Yes, the heroin example is quite extreme. The "war on drugs" is also quite extreme. For example prostitution is not legal in the Czech republic...but not strictly illegal either. That means the results are less severe than with the drug prohibition, but there are still bad consequences, I would imagine. Since the whole thing is quite shady, there are quite a few ukrainian prostitutes here who are illegals (another problem caused by regulation) and since they don't have a legal stay or a legal job their pimps have a lot of power over them.
Also drug prohibiton is less severe than it is in the US (or even than in Germany). Therefore the drugs tend to be cheaper and the information more plentiful. It is easier trace the drugs to the produced, therefore they are safer...still not as safe as they could be if they were legal.
So it is not binary, sure, there is always a scale, but any black market has some undesireable consequences compared to the situation where it is allowed to be a legal market.
@Tibor Mach:
I'm trying to point out that the sliding scale is not just based on the level of prohibition, but on the nature of the prohibited product and who tends to demand it. Look at heroin:
-There's moderate demand (IIRC fewer than 5% of marijuana users have tried heroin, and only a fraction of those who try heroin go on to use it regularly, and only a portion of those regular users become addicts).
-It's easy to transport and conceal; it is a lightweight powder that can be broken down into small packages or infused into other materials.
-It's only somewhat replicable with other substances; many prescription painkillers are used as substitutes by desperate users, some of them getting addicted to those instead.
-It can be used discreetly; say, alone in a bathroom stall with just a few fairly easily obtained paraphernalia or "works".
-It requires accomplices and interaction with others (unlike other proscribed activities such as, say, speeding, which you can do completely independently simply by pressing your foot further on the gas pedal).
-The people who use heroin tend to be seedy fringe types who are isolated from mainstream culture and lifestyles; these are people who are accustomed to violence and tend to be inarticulate and poorly connected to legitimate social infrastructures.
So the combination of all these characteristics means that a ban on heroin produces a violent and dangerous black market in heroin. Would that happen with technologically-assisted gene selection (TAGS), then? Let's compare it categorically:
-The demand is fairly low right now, but I think it's plausible that it could very quickly grow, since people overwhelmingly want smarter, healthier, stronger kids.
-I'm not sure what kind of equipment or tools are required to perform TAGS, but I'm guessing it's not stuff you can find lying around your house. The required raw data, on the other hand, is easy to transport and hide just like virtually all other modern data.
-The alternative to TAGS is natural eugenics, A.K.A. choosing mates with desirable genetic traits. This is basically what we do already, and what most of us seem perfectly happy doing.
-In theory, nobody has to know that your baby was illegally bred with TAGS, though I think there would always be some risk of that information getting out.
-Nobody has the ability to perform TAGS out of their own garage, without materials, expertise, etc. being sourced from somewhere else.
-The people who are most interested in--and probably most able to eventually afford--TAGS, probably overlap significantly with academic and futurist circles. Popular eugenics movements of the past were like this too (a few notable eugenicists: John Maynard Keynes, Sir Francis Galton). These are not the types who resort quickly to violence or life-threatening deceit. They tend to be familiar with, if not directly involved with, key instruments in our society's established mechanisms for handling disagreements. (David Friedman, for example, teaches classes on law.)
So, even if even minor TAGS infractions carried the most severe of penalties, I don't think there would be a dangerous black market, at least not if TAGS became both widely available and prohibited tomorrow.
The point of my "chunks of DNA" was that the process doesn't randomly select genes--genes that are close to each other on the strand are likely to stay together. So you don't get to select single genes in Heinlein's method.
But you do get to select a chunk that contains a gene you want, and to some extent, among such chunks, one that doesn't contain genes you don't want or does, if you are lucky, contain other genes you want.
Sigh.
I am well aware of how genomic chip technologies work (expression arrays, SNPs, oligos, etc.) having published several peer-reviewed papers that integrate these data, been awarded peer-reviewed grants based on such, and having personally read and reviewed dozens of relevant papers [1].
So I am well aware that "the process" -- assuming by that you mean GWAS -- is 1) not random in what it detects, but 2) most often DOES NOT detect a gene at all [2]. What gets detected as significantly associating with a trait is most often a single nucleotide (i.e. base -- A,G,C, or T) within the vast sea of DNA where there are no protein coding genes. These "variants" are referred to as "snips" (SNPs) -- Single Nucleotide Polymorphisms (plz see your 23andme report). Sometimes these snips fall within the protein coding region of a gene, not usually, but sometimes, and for some of those that do the polymorphism (compared to control) will change one amino acid in the cognate protein. These (non-synonymous snips) are the most helpful to biologists seeking to untangle the molecular cause the variant has wrt the trait. For the vast majority of other snips, at this point, even though they're statistically significant, nobody knows why.
So, pointing out that the process doesn't "randomly detect genes", while true -- it usually doesn't detect genes at all! -- doesn't really help you; it is not a suitable technology to guide the serial selection of "genes" that associate with a trait, be that Heinlein's technology or any other, and I hope by now it is clear to you why.
What part is unclear? The "precise set"?
Well... yes. Say you discover 25 variants highly significant for trait X. Because you've thrown biological function to the wind, you have no clue, not even an hypothesis, as to which variants are really important to trait X, so you have to isolate them all in the haploid genome of sperm/egg (as the case may be) in order to get them in the child's genome so she'll have chance of expressing trait X. Clever sequencing of one of four spermatids produced from one progenitor cell, without destroying it, to see which, if any, of the variants are present, doesn't achieve that -- you'd need to get all the variants in a single spermatid. Coupled with the problems you now acknowledge with mom's DNA, the entire approach is "unlcear", and I'm being generous.
But not when responding to an argument that was explicitly about the sperm and egg, not about other cells. If you followed the description of the technology,
But you see, that's the point, I wasn't "following" it. By that point in the thread I was trying to tease out your understanding of it, and from all that's been written here now I don't think you understand it very well, abstractly perhaps, but not at the level of what might even be feasible in terms of cell biology.
your answer made no sense unless you really believed what you wrote--that all the husband's sperm were identical and all the wife's eggs.
For cryin' out loud, leave it go, would ya David?
[1] Curious readers may go to my blog to obtain a list of references to my papers, to be found somewhere in the 'About Me' link.
[2] Only ~1.5% of the human genome codes for protein, what biologists nominally refer to as a gene.
@Tibor Mach:
PS. When I mentioned that people who use heroin tend to be fringe types who are accustomed to violence, I neglected to address the more important fact that people who supply heroin also tend to be fringe types who are accustomed to violence!
New York City has strict rules about the specifications of shower heads you may install. Outside of "Seinfeld," do you suppose there's a dangerous black market for shower heads on the streets of Manhattan?
Power Child: There is a reason this is true and that the same is not true about legal drugs such as alcohol. It is costly to do something that is illegal and persecuted to the point that it is called a "war". If you are already a unsavory type who has several other criminal activities, the costs are not that high. So these types dominate there. The same reason that during the alcohol prohibition in the US, it was the Italian mafia that took over most of the production.
The shower idea is what I tried to explain in the last comment. Here both the persecution and the demand are lower. Therefore the consequences are much much less drastic, they are rather a nuisance. But still they are negative.
Power Child: Sorry, I noticed your second comment first, so here is a reply to the first one:
Of course, demand is very important. If something is illegal, but noone really cares about that then there will not be a black market. If people care about it a little but the punishments are draconic, the black market prices would be too high to meet the demand. But I think people want this even more than (some people) want herion.
Other than that - I did not suggest illegality of this kind of treatments would fuel violence. Violence is not the only negative consequence of the black market. It is true most of those other negative consequences only affect the people who actually participate in those transactions. Someone could see that even as a good thing - as an additional detterence. But let's say nearly everyone would actually want to have "better" kids (as I would very much expect). Still, the state punishes any practicioner who would do something like that and also the customers as well. That raises the cost of the treatment, depending on the punishment possibly quite a lot. But isn't that exactly what the objection against that was in the first place? That the rich people would then simply have much more able children than the poor people? The ban would make the difference even more significant. Or, since the demand would presumably be very high, there could be a cheaper (and more dangerous) version provided as well. I'm thinking along the lines of crack vs. pure cocaine. Crack is a very dangerous drug...and it is cheap. Cocaine is of course something one has to be very responsible with but it is far less dangerous than crack. The whole existence of crack is the result of the drug prohibition as apart from the artificial costs of illegality, cocaine can be produced cheaply. War on drugs drives the price up and crack cocaine is then provided as a cheap alternative. I would expect something similar to emerge in the eugenics black market as well.
RKN: You don't need all 25 to increase the chances. 25 are significant. So of course, your chances of getting the desired trait are best if you have all of them present. That may turn out to be in conflict with some other traits you would like to see in the baby (as David pointed out - you make the best the enemy of the good). So instead you lower the chances a bit by selecting 5 instead...hoping that in fact these 5 are really the relevant ones (or that among them are the relevant ones). You play a gamble, sure. Then you observe the results. It turns out perhaps that these 5 were the wrong ones. Well, that changes your posterior distribution quite a bit. Now you are wiser and have improved the odds for the next generation. All of that without any theory or hypothesis of why that is so. I really don't understand what is difficult about that.
It just seems to me that you either want the improvements to happen with certainty which of course I completely agree that this process cannot give. But if someone told me in a casino that now I have 20% better odds (which still means I can lose of course), I would be quite happy about that (in fact in a casino, such a huge increase of chance would lead to a situation where I could ruin the casino in the long term in quite a few games, but never mind, this is entirely off topic). And the odds can get better in time percisely by observation that in the cases where some variants were chosen and some other were not present, not significant improvement happened, which is a reason to assume that those were probably not the right ones. So each generation, this technology would give better results even if the biological analysis gets stuck at the same level of understanding.
"I am well aware of how genomic chip technologies work"
I'm not, but that wasn't what I was talking about. I was talking about the process that produces a sperm or egg with half the full complement of genes.
"So, pointing out that the process doesn't "randomly detect genes"
I said nothing at all about detecting genes, randomly or otherwise, despite your using the word in quotation marks. My word was "select." I was talking about how, after deciding that you wanted certain genes because you thought they were related to desirable traits, you could actually get them by choosing which sperm and egg to combine. Your ability to do that is limited both by the limited number of sperm and (especially) eggs and by the fact that if two genes are very close to each other you probably won't get a sperm or egg that has one and doesn't have the other.
Is that clearer? Your post is mostly responding to your misunderstanding of what I was saying, whether because I did not write clearly enough, you did not read carefully enough, or some combination thereof. My point in what you responded to was not about the first stage of deciding what genes you think you want but about the seconds stage of, so far as practical, getting them.
"Say you discover 25 variants highly significant for trait X. Because you've thrown biological function to the wind, you have no clue, not even an hypothesis, as to which variants are really important to trait X, so you have to isolate them all in the haploid genome of sperm/egg (as the case may be) in order to get them in the child's genome so she'll have chance of expressing trait X. "
In order that she will have a *chance* of expressing that trait? That doesn't follow at all. You have just pointed out that I don't know which are important. If I end up missing an unimportant one, she still gets the trait. You seem to be insisting that one can't do anything unless one does it perfectly.
And I do have clues--because some will correlate more closely with the trait than others.
RKN: You don't need all 25 to increase the chances. 25 are significant. So of course, your chances of getting the desired trait are best if you have all of them present. That may turn out to be in conflict with some other traits you would like to see in the baby (as David pointed out - you make the best the enemy of the good). So instead you lower the chances a bit by selecting 5 instead...hoping that in fact these 5 are really the relevant ones (or that among them are the relevant ones). You play a gamble, sure. Then you observe the results. It turns out perhaps that these 5 were the wrong ones. Well, that changes your posterior distribution quite a bit. Now you are wiser and have improved the odds for the next generation. All of that without any theory or hypothesis of why that is so.
So you're prepared to caution the would-be parents thusly:
"Now, this is somewhat of a crap shoot -- Mr. and Mrs. James, is it? Yes, well, I am here to quell your concerns Mr and Mrs James. You see, by using our company's novel approach we can guarantee, with a 20% Bayseian probability, that your child will play the flute well. However, the genomic changes we'll need to make may affect other biological functions, we don't even pretend to know all the downstream, biological effects of the changes we make. Consequently, certain 'incidental' characteristics may arise in your child, for instance, she may well exhibit serious glucose uptake issues (that happened to the child of one of our clients so they put her up for adoption), she could become pre-diabetic before age 5, but hey, you did say your dream is to raise a flute player. Certainly you don't want to make the ideal the enemy of the good!"
I really don't understand what is difficult about that.
I mean this in all sincerity: I hope that if you do in fact go to work with a team of molecular biologists, geneticists in particular, you listen closely and learn. Then think back on this dialog and what I've said and see if you don't agree with all of it and understand why it is so difficult.
And this coming from someone whose present area of specialty is computational network biology, a variant of systems biology, has a prior MS degree in geophysics with 20+ yrs industrial experience modeling high-dimensional problems in the oil & gas business, an unrecognized minor in mathematics and a boat load of scientific software development experience. (Weird pedigree, I know).
So I am by no means cynical toward computational approaches. Quite the opposite. But molecular biology is hard, real hard, as I trust you will find out for yourself one day.
@DF
Fair enuf, when you wrote ...
The point of my "chunks of DNA" was that the process doesn't randomly select genes--genes that are close to each other on the strand are likely to stay together. So you don't get to select single genes in Heinlein's method.
... I understood "the process" to mean the upstream process of variant detection that gets the list of variants to guide the sperm/egg selection process.
Is that clearer?
Yes, that's clear.
But what remains unclear, and what was the likely source of my confusion, is why you continue to misuse the word gene? You got all over me for a *single* misstatement, using "sperm" to refer to "germ cell", a misstatement I acknowledged (twice), accounted for, and yet *you* continue to say "genes" when in fact the upstream process of variant detection doesn't detect genes, and only rarely detects variants within a gene. Why do you continue to misuse the term? And if you don't understand *that*, why should anyone trust that you understand the more complicated stuff?
In order that she will have a *chance* of expressing that trait? That doesn't follow at all. You have just pointed out that I don't know which are important. If I end up missing an unimportant one, she still gets the trait. You seem to be insisting that one can't do anything unless one does it perfectly.
And you seem to harbor a very naive view of genetics and complex traits if you think making the kinds of changes you propose to get the traits you want will cause no unwanted biological effects, even with the vanishingly small chance it would produce the trait you do want.
So you don't know which variants are essential to the trait, you've no clue what incidental downstream effects may result from changing any of them, so you take the, what, 50% most significant variants and give it spin in a real human "experiment?"
I await your reply as to what I didn't understand this time, must be something, because you'd have to be mad to suggest such a thing.
However, the genomic changes we'll need to make may affect other biological functions, we don't even pretend to know all the downstream, biological effects of the changes we make.
and
And you seem to harbor a very naive view of genetics and complex traits if you think making the kinds of changes you propose to get the traits you want will cause no unwanted biological effects, even with the vanishingly small chance it would produce the trait you do want.
I don't think you're being fair. You seem to think that deliberate changes to the genome will be both large (in that they will have a high chance of hurting someone) and small (in that the benefit will be vanishingly unlikely). We know that large chunks of chromosomes are recombined during meiosis, and are thus, potentially viable humans. Now why would you assume that the statisticians and geneticists will pick only the most likely modifications to do harm and least likely to be of benefit?
In responding this way, I'm not ignoring the point you made about the difficulties in detecting genes; but those are, in principle, limitations to technology that can be overcome. Whether the current technologies are too crude to be used at all in the genetic engineering that people described here is an empirical question, but likely can be answered in the negative, because eugenics is already in use to select fertilized eggs for implantation that don't carry certain easy to identify diseases.
@Tibor Mach:
I don't disagree with any of what you said. Yes, crack is to cocaine what moonshine is to wine or beer or what heroin is to opium.
Now imagine TAGS becomes widely available tomorrow, and illegal the next day after that.
Who do you suppose would be closer in persona to a typical seeker of illegal TAGS: David Friedman, or a corner street criminal? If you answer like me, you'll say "Much closer to David Friedman." So, how likely is a David Friedman to break the law and take associated risks just to get TAGS for his offspring? I say "Not very."
Obviously there will be a few who are, but by and large most will not, and so the black market will be relatively small and lacking in drama.
I don't think you're being fair. You seem to think that deliberate changes to the genome will be both large (in that they will have a high chance of hurting someone) and small (in that the benefit will be vanishingly unlikely).
Did I say or infer "large" in my example? I'll put it to you, Joe. How large does the statistical chance of running afoul an inadvertent pleiotropic effect of a genetic alteration have to be before you say, this is unethical, we shouldn't run this experiment in a human being, more specifically, *your* would-be child?
For a similar reason I think the chance of having a complex trait expressed, using this method alone, would be vanishingly small because complex traits -- hell, even most simple traits -- result from the activities of many bio-molecules (not just gene products) interacting with each other in a regulated fashion to bring about a given phenotype. That's a biological fact.
Genomic methods (e.g. GWAS) give us clues as to how the phenotype may be mediated at the level of DNA, but as I'm sure you're aware, there's a LOT more going on in the Proteome, to say nothing of environment, necessary to express a given phenotype.
Fair enuf?
In responding this way, I'm not ignoring the point you made about the difficulties in detecting genes; but those are, in principle, limitations to technology that can be overcome.
Sorry, Joe, I can't agree. There have been hundreds, probably thousands by now, papers published on GWAS results, obtained on HUGE cohorts and showcasing the coolest and most sophisticated Bayesian-markov-chain-monte-carlo-huma-huma methods for prediction ever invented. And yet, we're still scratching our heads.
As I said way up-thread somewhere, what we're talking about is not a TECHnology problem, it's a BIology problem. That may disappoint the future pundits among us, but so be it.
Whether the current technologies are too crude to be used at all in the genetic engineering that people described here is an empirical question, but likely can be answered in the negative, because eugenics is already in use to select fertilized eggs for implantation that don't carry certain easy to identify diseases.
I've already footnoted up thread the exception of diseases conferred by single gene mutations.
Look, I can tell you from firsthand experience, working in the area of biomarker discovery and companion drug diagnostics, that nobody, and I mean nobody, much cares how cool or statistically significant one's computational results are, everyone concerned will demand functional proof in one or more relevant biological models before they believe you. The bar is high.
I have to believe the bar for human eugenics of the kind discussed here would be -- should be -- much, much higher.
"However, the genomic changes we'll need to make may affect other biological functions, we don't even pretend to know all the downstream, biological effects of the changes we make."
Equally you don't know the effect of not making the "changes." All you are doing is selecting which sperm and which egg combine. There is no reason to think the ones you choose are in any respect worse than what would have happened by chance, some reason to think they are in some respects better--that being why you chose them.
"I await your reply as to what I didn't understand ..."
Happy to oblige.
You write:
"And you seem to harbor a very naive view of genetics and complex traits if you think making the kinds of changes you propose ..."
It sounds as though you still think I am making changes, somehow engineering the genes. Some day that may be practical. But all I am doing in Heinlein's approach is selecting, among the eggs and sperm produced, which egg combines with which sperm.
Did you understand that when you wrote your response? If so, can you explain why that selection imposes risks that random selection does not impose?
RKN:
Well, since your objections to both me and David seem to be basically the same, the answers are also. David's last comment (reply) is a reply to your James family scenario as well.
With perhaps one problem. There is one sperm that reaches the egg and I guess that even though it is probably (correct me if I am wrong, I claim no good knowledge of the relevant biology and definitely nowhere near your level) largely random, the abilities of each of them (again, probably) play a role as well. And so if fastest sperm is also in a way the healthiest sperm (on average, I am only assuming a correlation between being the first to get to the egg and inside and overall quality of it with respect to others in the particular ejaculate), your objection could be more valid.
Now you select a particular sperm for certain characteristics. As David pointed out, if you don't know that this one contains some bad stuff as well, you don't know the same about the others either. But if the bad information is less likely to be encoded in the likely winner of a natural fertilization it could be a potential problem. In short in the scenario as I described it it is possible, that natural selection could (under certain circumstances and relative probabilities and seriousness of the effects) in fact mean better chances.
Even if that is so, there are possible (partial) solutions to that problem.
First is to simply acumulate the sperms which contain what you want them to contain (and a lot of other information as well) and then let them compete the regular way. Assuming that the other information is i.i.d. (independet, identically distributed), you can get the benefit of both worlds.
Second, if even that does not work, you can still at least pick among the eggs. That is much more limited, but there is no correspondence with the possible "winning sperm" effect.
Third, you wait till the biological knowledge improves so that the advantages from artificial selection outweight those from selection by "winning the race".
By the way - If those assumptions about winning correlating with "good genes" are true, is a possible "eugenics" method just using a very large amount of semen during fertilization? Sorry if that is a stupid question...I guess the problem (even if otherwise the reasoning is correct, which might not be) is that the sperm which starts in the back simply has a too big a disadvantage that it basically cannot be compensated by higher speed.
Also, I will not be working this directly with biologists...I am working with branching and coalescent stochastic processes and they have a lot of applications in (among other things) geneaology and genetics.
[...] some reason to think they are in some respects better--that being why you chose them.
But the basis of this reason is biologically flawed, if you ask me anyway.
When I presented you the 25 variant scenario you accused me of making enemies of perfect and good. You said it wouldn't be necessary to get *all* the variants significant for a trait isolated in one sperm (egg), you seem to think some percentage of them should still increase the likelihood of the trait being expressed compared to none or fewer of the variants being present, but offhand I don't know of any biological evidence to support the notion that genomic variants of the kind we're talking about are additive toward producing traits in the way you evidently believe they are. That's one of the reasons why they're called "complex" traits. And why the problem isn't merely statistical or technological, it's fundamentally a biological problem. But I repeat myself.
Did you understand that when you wrote your response?
If you're asking did I fully understand the selection technology you were describing, no, not fully, not originally. (I might add I'm a tad skeptical you did either until ~75% or so into the thread, but I'll let that go).
I originally thought up-thread you were describing some process whereby one could (somehow) isolate in one sperm (egg) *all* the relevant variants detected as significant for some trait. But I see now it's merely to isolate from the population of all sperm in the ejaculate (?), by clever sequencing, the ones that contain some "percentage" of the relevant variants. Correct?
By the way, assuming this is correct, it just dawned on me, how do you propose to isolate the precise four spermatids and the correct diploid progenitor cell together? You need to do that to make the clever sequencing work, right? But the billions of spermatids are all mixed up in the ejaculate, and by then dissociated from their diploid progenitor?
If so, can you explain why that selection imposes risks that random selection does not impose?
Who told you natural selection of sperm/egg is random? Sperm (eggs) vary in their viability, owing, I suppose, to the quality of their chromatin (not my specialty). Intervening in the "natural" selection process in the way you propose to select sperm/egg pairs that are potentially less viable, certainly may lead to miscarriage or unexpected birth defects.
Btw, no reply to my criticism over your repeated misuse of the word "gene?"
I wrote:
... and the correct diploid progenitor cell together.
*head slap in shower*
Don't need them *together*; subtracting from the sequence of any diploid cell should do.
Rest of question stands.
you seem to think some percentage of them should still increase the likelihood of the trait being expressed compared to none or fewer of the variants being present, but offhand I don't know of any biological evidence to support the notion that genomic variants of the kind we're talking about are additive toward producing traits in the way you evidently believe they are.
Your skepticism is misplaced. If the genes that create something complex like musical ability were not, in some way, additive in their effect on fitness, they would simply not have evolved. Evolution rests on the idea that small alterations can change complex interactions.
If that weren't true, and say one would need to carry all of 10 gene variants to have a measurable increase in musical ability, and thus you would have to find the 1 in 2^10 sperm cell that carried all genes, then how would you explain the variety of musical abilities in siblings (or humans in general)?
I'll preempt what I think is an unsatisfactory answer to my question:
You could try to explain the variety of abilities among siblings by arguing that while you may need all of a set of 10 gene variants [genes A through J] to express a certain musical ability, there are many other ways to be musical (eg. gene variants K through Q; B through H etc).
This just means you would want to look at many different loci using Heinlein's technique; it doesn't have to reduce the effectiveness of the screen.
Your skepticism is misplaced. If the genes that create something complex like musical ability were not, in some way, additive in their effect on fitness, they would simply not have evolved. Evolution rests on the idea that small alterations can change complex interactions.
Your larger mistake, if you ask me, is in thinking that -- as you appear to -- every human phenotype is the result of neo-darwinian natural selection. Many behaviors in humans have no evidentiary basis in "reproductive success", e.g., exceptional musical ability, and in fact a good number of behaviors contradict it.
Even assuming that there is *always* a genetic basis for every phenotype, then how do you explain that when the alterations -- variants -- don't reside in the protein coding region of a gene, or the transcriptional control region of a gene (promoter, enhancer, etc.), that they still impact the complex interactions causing the phenotype?
If that weren't true, and say one would need to carry all of 10 gene variants to have a measurable increase in musical ability, and thus you would have to find the 1 in 2^10 sperm cell that carried all genes, then how would you explain the variety of musical abilities in siblings (or humans in general)?
The variation in musical ability across humans might very well be explained by variability in formative environment, a desire to learn, hours of practice, etc., rendering the 10 variants as nothing more than, well, associative.
And for the last time, call the variants what they ARE, genomic variants, not "gene" variants, something I'd hoped at least you understood. No evidence David does.
there are many other ways to be musical
No - you mean like hard work and practice!
This just means you would want to look at many different loci using Heinlein's technique; it doesn't have to reduce the effectiveness of the screen.
Except the Heinlein method (clever sequencing) won't work as described on sperm for the reason I mentioned up-thread. Unless someone tells me I'm mistaken. And we've already acknowledged the problem with eggs. So from a technological basis the whole thing is hooey.
Your larger mistake, if you ask me, is in thinking that -- as you appear to -- every human phenotype is the result of neo-darwinian natural selection.
You're right, I was being sloppy. But my argument still stands if you grant that some variation between siblings is influenced by genes. Which brings me to:
Even assuming that there is *always* a genetic basis for every phenotype, then how do you explain that when the alterations -- variants -- don't reside in the protein coding region of a gene, or the transcriptional control region of a gene (promoter, enhancer, etc.), that they still impact the complex interactions causing the phenotype?
1. "Always" isn't necessary, but I'll continue
2. I might be ignorant, but how does this weaken my argument? Noncoding areas of the genome also can be selected for.
The variation in musical ability across humans might very well be explained by variability in formative environment, a desire to learn, hours of practice, etc., rendering the 10 variants as nothing more than, well, associative.
Your argument requires that all of the variation comes from factors that are influenced by completely random events. The desire to learn may also be genetic.. in which case, Heinlein's technique can select it.
And for the last time, call the variants what they ARE, genomic variants, not "gene" variants, something I'd hoped at least you understood. No evidence David does.
I was speaking colloquially. All I said about gene variants can be interpreted in the "genomic variant" sense of the word. (Unless I'm misunderstanding something. See comment 2)
Your argument requires that all of the variation comes from factors that are influenced by completely random events.
Don't where you're getting that from. I acknowledge some complex traits may have a genetic basis, others not. I see no reason to call non-genetic bases "random."
The desire to learn may also be genetic.. in which case, Heinlein's technique can select it.
Heinlein's technique won't work --- see my 10/16 reply to David.
If it makes you happy, replace "random events" with "environment + random events".
But I am comfortable making the argument that if there is a relationship between environment and genes, it is most likely true that genes are the cause. Eg. animals with the genes that beavers have tend to live in dams. It's probably genes->live in dams, not the other way around.
Post a Comment