Physicist, Startup Founder, Blogger, Dad

Thursday, April 26, 2012

UC Irvine colloquum

I should have put this up earlier in case anyone in the LA area wanted to attend. But I was too busy at KITP and forgot. UCI Physics Colloquium (slides).
Genetics of Intelligence and Other Quantitative Traits

Date: Thu, 04/26/2012 Time: 3:30 pm Location: RH 101

How do genes affect cognitive ability or other quantitative traits such as height? I begin with a brief review of psychometric measurements of intelligence, introducing the idea of a "general factor" or IQ score. The main results concern the stability, validity (predictive power), and heritability of adult IQ. Next, I discuss ongoing Genome Wide Association Studies which investigate the genetic basis of intelligence. Due mainly to the rapidly decreasing cost of sequencing, it is likely that within the next 5-10 years we will identify genes which account for a significant fraction of total IQ variation.


dwbudd said...

Prof Hsu:

a very thoughtful and interesting talk. I wish I were still living out in Southern California.  Would love to hear you present it (looking at PP slides - a medium that is truly information limiting - only goes so far.)Two comments:

One, I think your layman's definition of "validity" misses the mark a bit.  Psychometrically, that a test is "predictable" is something of a surrogate for "validity," which means not that what you are measuring is "real," but rather, that what you purport to be measuring is, in fact, what you are measuring.  Simply put, imagine a scale that claims to weigh people, but instead captures how tall they are.  Height is "real," but the scale you are using is not really capturing it.  

Two, I wish Stephen Jay Gould were alive to see the outputs of your BGI project.  One of the maddening attacks on Charles Murray's The Bell Curve was that heritable intelligence does not exist because...well, it just can't.  Professor Gould, the most prominent of the critics, argued that because geneticists had not at the time identified any genetic markers on which intelligence could be identified, therefore one does not exist.  It's sophistry of course - the argument that because I have not found it yet, therefore it is not there.  But many seized on this pseudo-scientific "logic."

When (if?) such a genetic locus for intelligence is found, it's a shame that the good Professor will not be around to see it.

steve hsu said...

Perhaps it wasn't clear from the slides, but when I refer to predictive power of the test, I mean for real world outcomes (e.g., college GPA, income). Ability of the test to predict the outcome of a similar test at a later time is "reliability" not "validity".

liu89 said...

>Seek thousands of subjects with IQ +3 SD or higher (roughly 1in 1000).>US gifted education in last 20 years: SAT at age 12. Ceiling veryhigh: above 1 in 10,000.>We have obtained 2000 DNA samples from this 1 in 10,000population and will perform whole genome sequencing laterthis year.>Search for additional volunteers among this population underway – please volunteer!Do I understand you have already 2000 just at the 1 in 10,000 level? Could you say more about that sample? Also how many do you have at the level 1 in 1,000? And are you accepting more of those? And will they also get sequenced or only 1 in 10,000?

steve hsu said...

> Do I understand you have already 2000 just at the 1 in 10,000 level? Could you say more about that sample?
SAT administered to younger kids has a high ceiling. This has become the most common way to test gifted kids in the US.

> Also how many do you have at the level 1 in 1,000? And are you accepting more of those?

We set the threshold on our web site at roughly this level. You can volunteer there: http://www.cog-genomics.org

All accepted volunteers will get at least SNP genotyping, possibly whole genome sequencing.

BlackRoseML said...

 So a 700 V and a 800 Q GRE corresponds to a 145 IQ?

As for myself, I would suffer the indignity of being placed in the < 1280 SAT group (as I was highly neurotic and weary when I took the SAT, scoring low on the Math section), but my GRE results are mildly respectable, thus I prefer my intelligence to be evaluated based on my GRE scores.

Regarding the GRE: I tried to estimate my IQ based on my GRE results using simple linear extrapolation of the sum of the composite score. Since the US average is 1020 (table C1), I conservatively assumed that score corresponded to an IQ of 103, while a 1550 composite is a 145. My equation for this 0.0807x + 19.8 = IQ (where x is one's GRE composite score). Are there any critical flaws with this simple model or any reason why the relationship between IQ and GRE scores isn't linear? What is an appropriate IQ for the average domestic GRE examinee or the 1550 level? 

BlackRoseML said...


I found this lately: http://isteve.blogspot.com/2012/04/cochrans-new-theory-of-iq-genetics.html

So… most  genetic load in humans is made up of
many, many  mutations that each have fairly small effects.  A smaller
fraction of the genetic load consists of mutations with big effects on

... One important point
is that a single highly deleterious mutation has a good chance of
pushing the whole organism in some odd direction in phenotype space.  In
other words, the same mutation that drops your IQ, or damages your
heart, may also make you look funny.  At lower IQs, more and more kids
are considered to suffer from ‘organic’ retardation.  On the other hand,
a higher-than-average number of small-effect mutations should also
interfere with really complex systems such as the brain (and reduce IQ),
but because of the law of large numbers, wouldn’t tend to have any
particular direction in phenotype space.  As far as I can tell, an
extra-large dose of small-effect mutations, which we will henceforth
call genetic noise, would not make you funny-looking....If a kid’s parents have a higher-than-average
amount of genetic noise, on average the kid will as well. This sure
looks like what we usually call non-organic or familial retardation. Most
of the within-population variation in IQ looks to be familial rather
than organic.  If I’m right, this means that most IQ variation – what we
might call the normal range – is caused by differences in the number of
slightly deleterious mutations.  None of them would show up in a QTL
search, because all are rare. And that is where we stand thus far:  no
 intelligence QTLs have been found – although you never know what you’ll
see in the next population.  On the other hand, shared chromosomal
segments would mostly contain the same slightly deleterious mutations,
 and so IQ should correlate with genetic similarity, which is what
Visscher has found.

(Actually there are studies that demonstrate an inverse correlation between fluctuating asymmetry and intelligence.)

I would call this the Poisson model of IQ variation. For instance, let's assume, for a given population there are 100,000 loci and the deleterious minor allele would reduce one's IQ relative to the major allele (this reduction would be small enough so it would not be immediately selected out, either by severely impairing intellectual development or through other pleiotropic effects, but large enough to affect one's performance on intelligence tests). In a given population, the frequency of the deleterious allele would be small (say .3 %) and each loci is independent of other loci. Therefore, the distribution of deleterious alleles in a given person within the population would follow a Poisson distribution, with the mean burden of deleterious alleles being 300 (100k x .003) with the SD being ~17.3 (sqrt of 300 as the variance of a Poisson distribution is equal to the mean). For a six-sigma gulf, 100 deleterious alleles, this would be 1/1000th of the number of deleterious loci

Hmm... I wonder if genome sequencing would detect the genetic variants if Cochran's model predominantly explains a large proportion of intelligence variance. I suppose that if the deleterious allele exerts a large enough effect, it might be detected within families as siblings as less genetic variation among each other relative to the general population, thus it would be easier to detect the impact of one allele.

Blog Archive