RSS
One of the more intriguing current scientific papers is a recent one in Genetics with the unobtrusive title:
Detecting polygenic adaptation in admixture graphs
Fernando Racimo, Jeremy J. Berg and Joseph K. Pickrell
USA June 6, 2017
Abstract
An open question in human evolution is the importance of polygenic adaptation: adaptive changes in the mean of a multifactorial trait due to shifts in allele fre- quencies across many loci. In recent years, several methods have been developed to detect polygenic adaptation using loci identified in genome-wide association studies (GWAS). Though powerful, these methods suffer from limited interpretability: they can detect which sets of populations have evidence for polygenic adaptation, but are unable to reveal where in the history of multiple populations these processes occurred. To address this, we created a method to detect polygenic adaptation in an admixture graph, which is a representation of the historical divergences and admixture events relating different populations through time. We developed a Markov chain Monte Carlo (MCMC) algorithm to infer branch-specific parameters reflecting the strength of selection in each branch of a graph. Additionally, we developed a set of summary statistics that are fast to compute and can indicate which branches are most likely to have experienced polygenic adaptation. We show via simulations that this method – which we call PolyGraph – has good power to detect polygenic adaptation, and applied it to human population genomic data from around the world. We also provide evidence that variants associated with several traits, including height, educational attainment, and self-reported unibrow, have been influenced by polygenic adaptation in different human populations.
Self-reported unibrow (e.g., Leonid Brezhnev or Frida Kahlo, I presume) is a pretty funny phenotypic trait to investigate. I’m reminded of a throw-away line in Updike’s The Coup about a Soviet painting of “Brezhnev charming with the luxuriance of his eyebrows a flowery crowd of Eurasiatic schoolchildren.”
But some of the other traits Racimo et al investigate are more politically sensitive, enough to bring a lengthy caution in the same issue of Genetics:
Tread Lightly Interpreting Polygenic Tests of Selection
John Novembre*,1 and Nicholas H. Barton†
In this issue of GENETICS, a new method for detecting natural selection on polygenic traits is developed and applied to several human examples (Racimo et al. 2018). By definition, many loci contribute to variation in polygenic traits, and a challenge for evolutionary geneticists has been that these traits can evolve by small, nearly undetectable shifts in allele frequencies across each of many, typically unknown, loci. Recently, a helpful remedy has arisen. Genome-wide association studies (GWAS) have been illuminating sets of loci that can be interrogated jointly for changes in allele frequencies. By aggregating small signals of change across many such loci, directional natural selection is now in principle detectable using genetic data, even for highly polygenic traits. This is an exciting arena of progress – with these methods, tests can be made for selection associated with traits, and we can now study selection in what may be its most prevalent mode. The continuing fast pace of GWAS publications suggest there will be many more polygenic tests of selection in the near future, as every new GWAS is an opportunity for an accompanying test of polygenic selection. However, it is important to be aware of complications that arise in interpretation, especially given that these studies may easily be misinterpreted both in and outside the evolutionary genetics community. Here, we provide context for understanding polygenic tests and urge caution regarding how these results are interpreted and reported upon more broadly. …
Overall, the numerous complications described here, both technical and interpretative, are why this exciting field is still in its infancy. Progress is being made but most findings are wrapped in numerous caveats. For this reason, we caution that great care should be taken in communicating results of these studies to general audiences. Journalists producing simple headlines and/or taking results out of context have the potential to misconstrue the complexity and levels of uncertainty in an arena where simple misinterpretations come easily. Generally, authors of such studies, including Racimo et al., are cautious, and this degree of caution must not be lost in translation. This is particularly sensitive as polygenic selection studies are analyzing complex social and behavioral traits, and as social scientists look more keenly upon these studies as a source of inspiration for policy; these efforts may be premature, to say the least. We are in a time of extraordinary discoveries, but we must remember that even as we gain traction with new computational tools and expanded genomic studies, we still have a long way to go to make precise, fully supported statements about the nature of selection on complex traits in humans.
This seems like a pretty reasonable way to proceed.
OT Iran is now saying that it bribed Western leaders to sign on to the Iran Nuke Deal, and that if Trump is not successfully pressured back into it, then Iran will release the names of Western leaders who accepted bribes to cheerlead for the deal.
Leaving aside the fact that I feel like I just told a Mel Brooks joke (isn’t Iran now blowing up their own case? Even without the names, isn’t this own-foot-shooting?), here is an anonymous commenter’s bitter obsession with people who are more successful than he is:
With Kerry running around in Fuck the Logan Act mode the past two weeks, I don't see this as impossible.
I've yet to read a technical specification that detailed who was bribed.
https://usawatchdog.com/iran-nuke-deal-bribes-treason-and-fraud-dave-janda/
Interesting to contrast the appearance of recent American and Soviet/Russian leaders: Brezhnev had impressive eyebrows. Gorbachev was bald and had a great big winestain on his head. Yeltsin was missing a few fingers from one hand.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?
Perhaps there is something in superstition after all.
He picked up a live grenade when he was a kid. He used to apply butter to the stumps and eat the butter off.
"The government should be one fist!"
Fist envy?
He was a rotten premier and they had to get rid of him of course, but he could be funny.
https://ethicsalarms.files.wordpress.com/2016/07/crazy-hillary.jpg
On Mother's Day, it seems fitting that we should honor America's biggest *mother* by remembering key events in her life.
Many years ago, Hill and Bill saw the light:
https://www.youtube.com/watch?v=NWgeZCYYsqk
There was a Ukrainian Olympic gymnast who deliberately cultivated a unibrow to tap into ancient Eurasian steppes warrior-woman stylings (though it seems to me that a brow ridge might resemble a unibrow on primeval pottery). I tried to look up a picture for “ukrainian gymnast unibrow” and that was stupid of me.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?Replies: @snorlax, @Anonymous, @Anonymous, @Anonymous, @sayless, @Stan Adams
Trump and Obama are pretty interesting-looking.
One thing I will say in favor of the American press is they seem to be pretty good at catching a guy at the right angle to make him look like a chimp. They did it with Obama and Bush 43 both.
Is There Room in Diversity For White People?
http://quillette.com/2018/05/10/room-diversity-white-people/
OT NOT-SEES! NOT-SEES EVERYWHERE!
“Talk to your children about the Republican Party.”
I am being told that this is not satire.
Leaving aside the fact that I feel like I just told a Mel Brooks joke (isn't Iran now blowing up their own case? Even without the names, isn't this own-foot-shooting?), here is an anonymous commenter's bitter obsession with people who are more successful than he is:Replies: @Lugash, @Whiskey, @El Dato, @Intelligent Dasein, @Iberiano
“VAJA! I love VAJA! Tehran, if you’re listening, I’d love to see the rest of Kerry’s bank records!”
With Kerry running around in Fuck the Logan Act mode the past two weeks, I don’t see this as impossible.
Trump’s hair can’t be beat. Well, at least least since the last guy to wear a wig – James Monroe.
One thing I will say in favor of the American press is they seem to be pretty good at catching a guy at the right angle to make him look like a chimp. They did it with Obama and Bush 43 both.
I’m just thankful that a true genius like the Augustinian Friar, Fr. Gregor Mendel, picked simple peas 150 years ago to start the Mendelian genetics revolution.
Leaving aside the fact that I feel like I just told a Mel Brooks joke (isn't Iran now blowing up their own case? Even without the names, isn't this own-foot-shooting?), here is an anonymous commenter's bitter obsession with people who are more successful than he is:Replies: @Lugash, @Whiskey, @El Dato, @Intelligent Dasein, @Iberiano
The documents Bibi released were all technical. That’s why there were so many. Steve Wozniak was not in the loop Jobs deals if you get my drift.
I’ve yet to read a technical specification that detailed who was bribed.
Re: the commentary
I swear that with some effort one can find something exactly like that in the tone and the implied message published around 1948 in the Russian journal Доклады ВАСХНИЛ (Proceedings of the All-Union Academy of Agricultural Sciences; 1948 was when Lysenko’s faction achieved the final and officially certified victory over a few remaining geneticists.)
When I did a 23&me test, they have a whole ton of (voluntary) questions they ask, including lots of appearance questions; variations in unibrow-ness is one of the topics.
What, precisely, provoked that caution? Did the researchers mention intelligence and race?
And how silly to think that a science journalist, or blogger, or sub-editor who writes the headlines, is going to notice, and read, and understand, and read between the lines, and obey-yes-massa whatever the hell the caution is trying to say.
If they really want to change media coverage, they need to be concise, transparent, reference specifically what the danger is, explain pitfalls in plain English, offer a concrete worked example of how misinterpretation can occur, and give quotable sentences embodying the caution for use by writers. And do all this in half the word-count of what they wrote.
What they did is simply intramural CYA.
As the British golf commentator, Peter Alliss, put it, ‘You could get a pair of swifts to nest in those eyebrows ‘.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?Replies: @snorlax, @Anonymous, @Anonymous, @Anonymous, @sayless, @Stan Adams
As an old joke, attributed to that late British comedian, Stan Boardman, had it, ‘Gorbachev *does* want to take over the world – that’s why he’s got a map of it tattooed on his head’.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?Replies: @snorlax, @Anonymous, @Anonymous, @Anonymous, @sayless, @Stan Adams
No one of a superstitious bent would ever have allowed a man with such an obvious – and fearsome – blemish to assume absolute leadership.
Perhaps there is something in superstition after all.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?Replies: @snorlax, @Anonymous, @Anonymous, @Anonymous, @sayless, @Stan Adams
The emphysemiac Chernenko was described as being ‘quite frankly corpse-like’ by the late UK Labour politician Denis Healey. Incidentally, Healey himself sported a luxuriant uni-brow to rival Brezhnev’s.
He didn’t start the revolution. His work was overlooked for decades. He was very lucky that the people who rediscovered the phenomenon were such diligent and honest scholars that they found and acknowledged his work. (Many physicists, to take an example, probably would not have acted with such propriety.)
“Tread Lightly Interpreting Polygenic Tests of Selection”: the man on the gibbet treads very lightly indeed.
There can not be any true happiness or female empowerment until everything is a truly Islamic Zimbabwe with Mexican Roots tirelessly working for a tenured (((Establishment))).
(And then Matt Damon has to bring a last dose of freedom to the masses)
What or who do you mean?
I have a mathematician friend who, on starting to share a building with mathematical physicists, declared them all to be pushy, sharp-elbowed spivs whom he wouldn't trust an inch. I laughed because that reflected some of my experience as an undergraduate a million years ago. The greatest insult to which I have ever been subjected was after I had done particularly well in a physics examination and my tutor declared "you really are one of us". Not bloody likely.Replies: @utu
Leaving aside the fact that I feel like I just told a Mel Brooks joke (isn't Iran now blowing up their own case? Even without the names, isn't this own-foot-shooting?), here is an anonymous commenter's bitter obsession with people who are more successful than he is:Replies: @Lugash, @Whiskey, @El Dato, @Intelligent Dasein, @Iberiano
Sounds like random “news” from a neocon think tank or the Jerusalem Post, i.e. stuff for Murrican Consumer Bodies – hot on the heel of the Iran being fined for doing 9/11 etc. Link please.
I don’t remember any “cheerleading”. It was a hard slog for a completely unremarkable deal. Israel ran interference at every corner, and topped it up with assassinations for good measure.
Steve Hsu views the approach his group is taking and GWAS as distinctly (maybe fundamentally?) different.
———————
background:
My understanding of GWAS: get a bunch of DNA samples with associated phenoypes. Split them into binary groups based on phenotype (i.e. they have XYZ trait or they don’t). Then, across the genome, analyze the gene frequencies at each gene/SNP looking for disparities between the groups. The goal is to be able to say: this gene is unequivocally statistically associated with XYZ trait based on the frequency differences b/w the two groups.
What Steve’s group does: get a bunch of DNA samples with associated phenotypes. Run the DNA through some machine-learning type program, and based on the idea (theory?) of Compressed Sensing, one can compare small segments of DNA of various subgroups of the population– the computer basically mixes and matches these small segments of small groups for countless iterations and eventually gets associations to the phenotype.
—————
So it seems Hsu thinks that while GWAS has its uses, his way is hands-down more powerful/useful. Yes, his group’s predictor will come up with some SNPs that are false positives–but it’s bupkis in the scheme of things. You’ve got your damn predictor, after all. The same process can easily be run on numerous traits.
And furthermore, he’s using relatively complex math and computers while the GWAS folks are a bunch of plodders. Tortoise vs hair. He’s in the sexy, Smart Money Silicon Valley startup crew and GWAS is Old Economy Steve (pardon the unfortunate name confusion).
There’s seemingly a bit of hubris, but perhaps it’s warranted.
So I guess what I’m getting at: is my interpretation of this state of things correct? Would someone more knowledgeable care to expound on this?
This might be helpful: A tutorial on conducting genome‐wide association studies: Quality control and statistical analysis
https://onlinelibrary.wiley.com/doi/full/10.1002/mpr.1608
My (somewhat simplistic) view is as follows.
There are two aspects to GWAS: statistical tests of association (does this SNP matter in a statistically rigorous fashion?) and polygenic score (aka PGS, PRS) calculation (how big of a difference does this SNP make?). Hsu's work focuses on the latter. Note his company: https://genomicprediction.com/team/
More on that: http://infoproc.blogspot.com/2017/11/the-future-is-here-genomic-prediction.html
My understanding is that PGS derivation can be any of a number of model building techniques divided into two types. Either binary classification (for binary traits): https://en.wikipedia.org/wiki/Binary_classification
Or continuous variable prediction (for quantitative traits) which is often just called "regression" even though other techniques are possible: https://en.wikipedia.org/wiki/Regression_analysis
Focusing on prediction of continuous variables, I think most GWAS just use linear or logistic regression. Hsu's Compressed Sensing work is different because it uses CS which is also known as LASSO or L-1 penalized regression. https://en.wikipedia.org/wiki/Lasso_(statistics)
The key difference is that CS/LASSO tends to enforce a sparse solution. In other words, only a relatively small number of SNPs (say ~10,000 out of 1 million) end up being used. This appears to correspond well to reality and has theoretical and computational benefits.
I think the results they achieved for height speak for themselves: http://infoproc.blogspot.com/2017/09/accurate-genomic-prediction-of-human.html
I am very curious whether Compressed Sensing will become more popular in GWAS.
P.S. I think you are in the right ballpark for how things are being spun.Replies: @anon
Leaving aside the fact that I feel like I just told a Mel Brooks joke (isn't Iran now blowing up their own case? Even without the names, isn't this own-foot-shooting?), here is an anonymous commenter's bitter obsession with people who are more successful than he is:Replies: @Lugash, @Whiskey, @El Dato, @Intelligent Dasein, @Iberiano
I don’t see how. Lobbying is just bribery by another name, and why on earth wouldn’t Iran lobby for a deal that was good for them? This makes Iran look entirely reasonable in my opinion, and it reveals the duplicity of Western bureaucrats. I think that’s how the public at large will see it.
Leaving aside the fact that I feel like I just told a Mel Brooks joke (isn't Iran now blowing up their own case? Even without the names, isn't this own-foot-shooting?), here is an anonymous commenter's bitter obsession with people who are more successful than he is:Replies: @Lugash, @Whiskey, @El Dato, @Intelligent Dasein, @Iberiano
What pisses me off is, it’s bad enough US Taxpayers (read: Middle class) get forced to foot the bill for all sorts of welfare and government “programs” that support everyone but the descendants of the founders of this nation (i.e. millions of European-Americans), we get to also have have till shaved to provide, billions, drillions, quintoplehillions–whatever it may be–to support foreign nations directly.
https://usawatchdog.com/iran-nuke-deal-bribes-treason-and-fraud-dave-janda/
I always thought there was something vaguely Asiatic about Brezhnev’s appearance. But he’s apparently fully Russian.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?Replies: @snorlax, @Anonymous, @Anonymous, @Anonymous, @sayless, @Stan Adams
“Yeltsin was missing a few fingers from one hand”
He picked up a live grenade when he was a kid. He used to apply butter to the stumps and eat the butter off.
“The government should be one fist!”
Fist envy?
He was a rotten premier and they had to get rid of him of course, but he could be funny.
Are unibrow people reliable self-reporters? How many are in the closet, disguising themselves simply by plucking and tweezing? Possibly there’s more than meets the eye. They’ll never tell and we’ll never know.
The unibrow data are from 23&me.
Here is an example of Racimo’s findings showing positive selection in height for Europeans.
The corresponding graphics for the unibrow trait (e.g. figure S38) show selection against unibrow in Europeans.
This is an interesting look at allele frequency differences between Chinese and Europeans for all 3 traits Racimo studied.
More in this comment: https://www.unz.com/jthompson/piffers-equation-further-updated/#comment-2307038
And against all the prudent advice that the authors of the note give, here are the strongest results found by Racimo et al.:
Positive selection for loci associated with Educational Attainment in East Asians.
Positive selection for loci associated with Height in Europeans.
Negative selection for loci associated with Unibrow in Europeans.
Positive selection for loci associated with Photic Sneeze Reflex in Oceanians.
All the relatively high-effect SNPs associated with EA do not differ appreciably in frequencies between Europeans and Chinese.
See how truly scary this stuff can be? Always use abundance of caution! These genetic things are so complex that we likely will never be able to say anything completely unambiguous about genetic differences in complex traits between human populations.
The iffy ones I have encountered I am not going to name. If you want a public example look at, for instance, the perpetually grumpy, suspicious, ungenerous Newton, or at Einstein who seemed somewhat economical with the truth in acknowledging his intellectual forerunners in his Special Relativity paper – Poincare, for instance.
I have a mathematician friend who, on starting to share a building with mathematical physicists, declared them all to be pushy, sharp-elbowed spivs whom he wouldn’t trust an inch. I laughed because that reflected some of my experience as an undergraduate a million years ago. The greatest insult to which I have ever been subjected was after I had done particularly well in a physics examination and my tutor declared “you really are one of us”. Not bloody likely.
I did not have as bad experience with physicists as your friend had but I knew several who were a-holes I would not trust. I have a much better experience with mathematicians. But the precautions taken by Andrew Wiles indicate that things are not good there either.
Alexius Ducas, “discriminated by the epithet of Mourzoufle, which in the vulgar idiom expressed the close junction of his black and shaggy eyebrows,” says Gibbon, became Emperor in the Constantinople by poisoning and strangling his predecessor. He was deposed a short time later by the Latin crusaders in 1204.
I thought that the birthmark looked like Alaska, and it meant that he wanted it back.
Us unibrows are a much oppressed people otherwise, reparations and affirmative action are the minimum I will accept as recompense.
It is almost as if Chernenko and Andropov died quickly because they were boring-looking.
Who was the last interesting-looking American leader? Taft?Replies: @snorlax, @Anonymous, @Anonymous, @Anonymous, @sayless, @Stan Adams
I’ve always thought that Hillary looks interesting … from a clinical perspective:
On Mother’s Day, it seems fitting that we should honor America’s biggest *mother* by remembering key events in her life.
Many years ago, Hill and Bill saw the light:
Monobrow, Shirley?
Then there’s the ticklish brush with matters intellectual: highbrow or lowbrow?
Agree. The depraved lust many physicists have for validation as the smartest guy ever, or some subset thereof, is both astonishing and sickening.
Positive selection for loci associated with Educational Attainment in East Asians.
Positive selection for loci associated with Height in Europeans.
Negative selection for loci associated with Unibrow in Europeans.
Positive selection for loci associated with Photic Sneeze Reflex in Oceanians.
All the relatively high-effect SNPs associated with EA do not differ appreciably in frequencies between Europeans and Chinese.
See how truly scary this stuff can be? Always use abundance of caution! These genetic things are so complex that we likely will never be able to say anything completely unambiguous about genetic differences in complex traits between human populations.Replies: @Steve Sailer
Thanks. I’ve got the Photic Sneeze Reflex pretty strongly — I look at the sun and sneeze. Must be some overlooked Polynesian ancestry!
---------------------
background:
My understanding of GWAS: get a bunch of DNA samples with associated phenoypes. Split them into binary groups based on phenotype (i.e. they have XYZ trait or they don't). Then, across the genome, analyze the gene frequencies at each gene/SNP looking for disparities between the groups. The goal is to be able to say: this gene is unequivocally statistically associated with XYZ trait based on the frequency differences b/w the two groups.
What Steve's group does: get a bunch of DNA samples with associated phenotypes. Run the DNA through some machine-learning type program, and based on the idea (theory?) of Compressed Sensing, one can compare small segments of DNA of various subgroups of the population-- the computer basically mixes and matches these small segments of small groups for countless iterations and eventually gets associations to the phenotype.
---------------
So it seems Hsu thinks that while GWAS has its uses, his way is hands-down more powerful/useful. Yes, his group's predictor will come up with some SNPs that are false positives--but it's bupkis in the scheme of things. You've got your damn predictor, after all. The same process can easily be run on numerous traits.
And furthermore, he's using relatively complex math and computers while the GWAS folks are a bunch of plodders. Tortoise vs hair. He's in the sexy, Smart Money Silicon Valley startup crew and GWAS is Old Economy Steve (pardon the unfortunate name confusion).
There's seemingly a bit of hubris, but perhaps it's warranted.
So I guess what I'm getting at: is my interpretation of this state of things correct? Would someone more knowledgeable care to expound on this?Replies: @Steve Sailer, @res
Thanks.
Hey, Yasir Arafat tied his shawl-thingy just so, and it laid in the shape of prewar Palestine on his back.
No doubt he had the classic Sintashta/steppe genotype, if not phenotype.
---------------------
background:
My understanding of GWAS: get a bunch of DNA samples with associated phenoypes. Split them into binary groups based on phenotype (i.e. they have XYZ trait or they don't). Then, across the genome, analyze the gene frequencies at each gene/SNP looking for disparities between the groups. The goal is to be able to say: this gene is unequivocally statistically associated with XYZ trait based on the frequency differences b/w the two groups.
What Steve's group does: get a bunch of DNA samples with associated phenotypes. Run the DNA through some machine-learning type program, and based on the idea (theory?) of Compressed Sensing, one can compare small segments of DNA of various subgroups of the population-- the computer basically mixes and matches these small segments of small groups for countless iterations and eventually gets associations to the phenotype.
---------------
So it seems Hsu thinks that while GWAS has its uses, his way is hands-down more powerful/useful. Yes, his group's predictor will come up with some SNPs that are false positives--but it's bupkis in the scheme of things. You've got your damn predictor, after all. The same process can easily be run on numerous traits.
And furthermore, he's using relatively complex math and computers while the GWAS folks are a bunch of plodders. Tortoise vs hair. He's in the sexy, Smart Money Silicon Valley startup crew and GWAS is Old Economy Steve (pardon the unfortunate name confusion).
There's seemingly a bit of hubris, but perhaps it's warranted.
So I guess what I'm getting at: is my interpretation of this state of things correct? Would someone more knowledgeable care to expound on this?Replies: @Steve Sailer, @res
I think this is a pretty good summary:
But I think you are lumping together GWAS for binary traits and GWAS for continuous (quantitative) traits.
This might be helpful: A tutorial on conducting genome‐wide association studies: Quality control and statistical analysis
https://onlinelibrary.wiley.com/doi/full/10.1002/mpr.1608
My (somewhat simplistic) view is as follows.
There are two aspects to GWAS: statistical tests of association (does this SNP matter in a statistically rigorous fashion?) and polygenic score (aka PGS, PRS) calculation (how big of a difference does this SNP make?). Hsu’s work focuses on the latter. Note his company: https://genomicprediction.com/team/
More on that: http://infoproc.blogspot.com/2017/11/the-future-is-here-genomic-prediction.html
My understanding is that PGS derivation can be any of a number of model building techniques divided into two types. Either binary classification (for binary traits): https://en.wikipedia.org/wiki/Binary_classification
Or continuous variable prediction (for quantitative traits) which is often just called “regression” even though other techniques are possible: https://en.wikipedia.org/wiki/Regression_analysis
Focusing on prediction of continuous variables, I think most GWAS just use linear or logistic regression. Hsu’s Compressed Sensing work is different because it uses CS which is also known as LASSO or L-1 penalized regression. https://en.wikipedia.org/wiki/Lasso_(statistics)
The key difference is that CS/LASSO tends to enforce a sparse solution. In other words, only a relatively small number of SNPs (say ~10,000 out of 1 million) end up being used. This appears to correspond well to reality and has theoretical and computational benefits.
I think the results they achieved for height speak for themselves: http://infoproc.blogspot.com/2017/09/accurate-genomic-prediction-of-human.html
I am very curious whether Compressed Sensing will become more popular in GWAS.
P.S. I think you are in the right ballpark for how things are being spun.
I have always felt rage at the sound of people eating. Thanks to UR I now know the name for the condition.
I have a mathematician friend who, on starting to share a building with mathematical physicists, declared them all to be pushy, sharp-elbowed spivs whom he wouldn't trust an inch. I laughed because that reflected some of my experience as an undergraduate a million years ago. The greatest insult to which I have ever been subjected was after I had done particularly well in a physics examination and my tutor declared "you really are one of us". Not bloody likely.Replies: @utu
Newton was pretty awful person. Einstein was more charming. I have written several comments about it recently. Also there was a case of Einstein lifting whole paper of Klein in 1927 and basically being caught. Why did he do it then? Did not have to. A sign of some personality disorder? You can find my comments searching archives for Newton, Einstein or Klein.
I did not have as bad experience with physicists as your friend had but I knew several who were a-holes I would not trust. I have a much better experience with mathematicians. But the precautions taken by Andrew Wiles indicate that things are not good there either.
Me, too, but I’ve always attributed it to my Swedish genes. It runs strongly in my family with four generations in that line exhibiting it. For what it’s worth, somewhat counterintuitively I also find looking into light a good way to *suppress* an unwelcome sneeze — a trick taught me by my Swedish grandmother.
This might be helpful: A tutorial on conducting genome‐wide association studies: Quality control and statistical analysis
https://onlinelibrary.wiley.com/doi/full/10.1002/mpr.1608
My (somewhat simplistic) view is as follows.
There are two aspects to GWAS: statistical tests of association (does this SNP matter in a statistically rigorous fashion?) and polygenic score (aka PGS, PRS) calculation (how big of a difference does this SNP make?). Hsu's work focuses on the latter. Note his company: https://genomicprediction.com/team/
More on that: http://infoproc.blogspot.com/2017/11/the-future-is-here-genomic-prediction.html
My understanding is that PGS derivation can be any of a number of model building techniques divided into two types. Either binary classification (for binary traits): https://en.wikipedia.org/wiki/Binary_classification
Or continuous variable prediction (for quantitative traits) which is often just called "regression" even though other techniques are possible: https://en.wikipedia.org/wiki/Regression_analysis
Focusing on prediction of continuous variables, I think most GWAS just use linear or logistic regression. Hsu's Compressed Sensing work is different because it uses CS which is also known as LASSO or L-1 penalized regression. https://en.wikipedia.org/wiki/Lasso_(statistics)
The key difference is that CS/LASSO tends to enforce a sparse solution. In other words, only a relatively small number of SNPs (say ~10,000 out of 1 million) end up being used. This appears to correspond well to reality and has theoretical and computational benefits.
I think the results they achieved for height speak for themselves: http://infoproc.blogspot.com/2017/09/accurate-genomic-prediction-of-human.html
I am very curious whether Compressed Sensing will become more popular in GWAS.
P.S. I think you are in the right ballpark for how things are being spun.Replies: @anon
Thanks for your reply. I have some weekend reading to look forward to.