Sunday, August 18, 2013

Dreams of DNA machines

I'm often asked about the status of the BGI cognitive genomics project. A fairly long article about it appeared in WIRED recently. I have quite a lot of experience with the media from tech startup days, as well as from coverage of my physics and genomics research. My advice is to read everything with a grain of salt. Most journalists have the best intentions, but the complexity of the topics they try to cover makes their task extremely difficult.

While I cannot give a comprehensive update, I can state that
1. Volunteers who qualified for the study via and returned their samples by approximately September 2012 have all been sequenced on the Illumina platform, and will receive updates on the status of their genomic data relatively soon.

2. The relationship between BGI and Illumina has deteriorated since the acquisition of Complete Genomics by the former, which was vigorously opposed (ostensibly on national security grounds, believe it or not) by the latter. Our project has not escaped collateral impact from this development.

In other news, the documentary DNA Dreams (about our project) has won the Film & Science Award and has been selected by various film festivals throughout the world, including in Italy, France, USA and Denmark and the Grand Competition of Pariscience. It has also been acquired by several international broadcasters, including Sweden (SR), Japan (NHK), Germany and France (ARTE). See trailer below.

DNA dreams.

My comments on the documentary from an earlier post:
1. As you might expect, it emphasizes sensational aspects of our research -- genetic engineering, drugs for cognitive enhancement, etc. These are all possibilities, obviously topics we discussed at the behest of the film makers, but of course for now our work is basic research with no near term applications. (Basic research tends to be less interesting to viewers than science fiction extrapolations.)

2. I find the video visually interesting, but at times it emphasizes the alien or sinister. Even the musical background seems chosen for this purpose.

3. Several important members of our team have little or no role in the documentary, despite being interviewed extensively during its making. I suppose the director was limited in what she could include, given the 60 minute format. The young woman who leads the cloning team is not actually part of our group.


Diogenes said...

Steve mentions the "longest lived humans". This is certainly where I would use limited resources, but there is an interesting point here.

Long-life does apparently run in some families, but the correlation of age at death of identical twins is...0 and most of the very long lived did not have especially long-lived parents or relatives.

steve hsu said...

I've never looked carefully at heritability of longevity, but I thought it was at least moderately heritable -- like 25 or 30 percent of variance.

Diogenes said...

I would have expected the same too, especially considering the heritability of BMI, heart disease, etc. and that in general smaller individuals within a species of mammal live longer and height and "robustness" are very heritable.

My primary source is a text by Caleb Finch of USC. I looked it up. It was Chance, Development, and Aging, 2000 or earlier.

One problem which biologists might not catch is that age at death is not normally distributed, so the twin-twin distribution will not be bivariate normal and therefore the Pearson rho will be an inappropriate measure of correlation.

LondonYoung said...

Is the idea that Pearson corr of lifespans looks very different from Pearson corr of lifespans contingent on reaching, say, age 70?

David Coughlin said...

How would you define "the correlation for age at death of identical twins" ?

Richard Seiter said...

Interesting (and also not what I would have expected). I believe the data you cite for humans is from (but that does not contain the scatter plot, which seems odd). The data is from Danish twins born from 1870-1900 so I wonder if two world wars might have an impact. It would be interesting to see some analysis of particular cases (say twins with more than a 20 year difference in longevity, note they did exclude cases where a twin died before 15).
Finch quoted the heritability as 1% narrow sense, 23% broad sense.

ben_g said...

Thats incredibly surprising. Even if the correlation is more like 20% (there has to be some correlation given other factors that correlate with both genes and age), it's surprising.

What's the effect of shared environment vs nonshared environment/noise?

Has huge implications for our understanding of nutrition and health.

ben_g said...

how much variation in one twins age at death predicts variation in anothers?

David Coughlin said...

Can you formulate that for me?

ben_g said...

Check out Purcell's behavior genetics, some of it online here

Basically you're talking about rMZ which is the intrapair correlation..

Not sure how you'd address the lack of a normal distribution, etc. been a while since i did any of this stuff

gacl said...

I'd love to hear more details about the project and ideas for others interested in genetics. Documentaries deliver entertainment and focus on superficial novelty. The blogs of scientists should deliver more substance.

Diogenes said...

"life-style" may be too complicated for any hard science.

The one fact which, for me, towers above all others in this field, if it really is a fact:

Primitives, even when they live to great ages, do not die from atherosclerosis or cancer, etc.

From quite a young age, twelve, I've been acutely aware of the vanity of "modern life". Now the third, developing, world is close to the first in life expectancy. It's most of what's made me a Cynic.

Diogenes said...

There is a restriction of range here, right?

The locus of the marginal means is the straight line y = x, so Pearson isn't uninformative.

I'd have to read up on it.

DK said...

I don't have a lot of experience with media but it seems to me that so far the hype/output ratio is very exceptionally high. IIRC, you said previously that the first results are expected this summer. In the context, "results" did not mean sequencing data (no one ever doubted that part of the project). It is perfectly understandable if you cannot talk about things that are too preliminary. Still, a fair question: while you are getting some work done (however large or small it is), are you guys getting any results?

Emil Kirkegaard said...

That can't be right, and I googled it, and it wasn't right.



The aim of this
study was to explore, in a large and non-censored twin cohort, the
nature (i.e., additive versus non-additive) and magnitude (i.e.,
heritability) of genetic influences on inter-individual differences in
human longevity. The sample comprised all identified and traced
non-emigrant like-sex twin pairs born in Denmark during the period
1870-1900 with a zygosity diagnosis and both members of the pairs
surviving the age of 15 years. A total of 2872 pairs were included. Age
at death was obtained from the Danish Central Person Register, the
Danish Cause-of-Death Register and various other registers. The sample
was almost non-censored on the date of the last follow-up (May 1, 1994),
all but 0.6% had died, leaving a total of 2872 pairs for analysis.
Proportions of variance attributable to genetic and environmental
factors were assessed from variance-covariance matrices using the
structural equation model approach. The most parsimonious explanation of
the data was provided by a model that included genetic dominance
(non-additive genetic effects caused by interaction within gene loci)
and non-shared environmental factors (environmental factors that are
individual-specific and not shared in a family). The heritability of
longevity was estimated to be 0.26 for males and 0.23 for females. The
small sex-difference was caused by a greater impact of non-shared
environmental factors in the females. Heritability was found to be
constant over the three 10-year birth cohorts included. Thus, longevity
seems to be only moderately heritable. The nature of genetic influences
on longevity is probably non-additive and environmental influences
non-shared. There is no evidence for an impact of shared (family)

Lower than I expect modern data to show. See also:

Diogenes said...

"That can't be right, and I googled it, and it wasn't right."

It is right emil. You're source is wrong. See comments below.

Diogenes said...

Right. I think that's wrong. rho squared for MZ twins is an upper bound for broad heritability. A higher figure is meaningless.

And it is only an upper bound, because the MZ twins may share environments/environmental factors.

It's very simple. When all residual variance must be due to environmental differences rho squared is the fraction of total variance explained by genes.

Diogenes said...

And you should know that such small heritabilities are very likely meaningless even if correctly derived (and in this case they were derived incorrectly).

Quantitative traits may conceal discrete traits. Here's an example:

Extent of cerebellar atrophy and consequent ataxia (as measured by ICARS) are quantitative traits and are correlated with duration and heaviness of alcohol consumption for the population, yet at least half of very very heavy drinkers are immune.

I remember in the 90s those who'd survived AIDS, ARC, or HIV+ for a decade or longer were dismissed by sophistical (though they though themselves sophisticated) medical scientists as anomalies to be "expected" given a distribution of years to death from onset. They made the mistake of thinking of distributions as inhering in the thing in itself rather than as a mere tool.

And they were subsequently proven to be fools. Western Europeans are uniquely resistant to HIV disease.

Richard Seiter said...

I think the focus on distributions rather than attempting to identify subpopulations with different outcomes (and then finding out why and if it is possible to treat someone so they move to a lower risk population) is a major shortcoming in much current medical research. The genomics revolution is a good step in addressing this, but I wonder if adding metabolomics will be the real revolution. IMHO what medicine needs is some good, cheap, easily obtainable (e.g. non-invasive) biomarkers to correlate with health status and outcomes. (the main focus of this right now seems to be cholesterol, but having looked at an all cause mortality rate chart for cholesterol I question the wisdom of the approach being taken: )

Was anyone able to determine why some of the heavy drinkers were immune?

Diogenes said...

I don't know. I couldn't find any. But there have been cases of alcoholic ataxia developing in people who don't drink very much at all. So it seems there is continuous variation after a "tipping point".

When h^2 is 0 and H^2 is claimed to be positive, I dismiss it. All of the effect is due to epistasis, dominance, the womb? Alleles make no difference? Really?

And I made that mistake above. I should have said, "There is a sub-population of Western Europeans who are resistant to HIV disease."

Diogenes said...

Here's a link

Blog Archive