All hail Khan!

by on June 14, 2014 at 10:58 am in Current Affairs, Law, Religion, Science | Permalink

Congratulations to Razib Khan, the noted genetics blogger, on the birth of his son. Born just last week, Razib’s son is already making the news:

An infant delivered last week in California appears to be the first healthy person ever born in the U.S. with his entire genetic makeup deciphered in advance.

Razib, a graduate student at a lab at UC Davis in California, had some genetic material from his in-womb son from a fairly standard CVS test.

When Khan got the DNA earlier this year, he could have ordered simple tests for specific genes he was curious about. But why not get all the data? “At that point, I realized it was just easier to do the whole genome,” he says. So Khan got a lab mate to place his son’s genetic material in a free slot in a high-speed sequencing machine used to study the DNA of various animal species. “It’s mostly metazoans, fish, and plants. He was just one of the samples in there,” he says.

The raw data occupied about 43 gigabytes of disk space, and Khan set to work organizing and interpreting it. He did so using free online software called Promethease, which crunches DNA data to build reports—noting genetic variants of interest and their medical meaning. “I popped him through Promethease and got 7,000 results,” says Khan.

Promethease is part of an emerging do-it-yourself toolkit for people eager to explore DNA without a prescription. It’s not easy to use, but it’s become an alternative since the FDA cracked down on 23andMe.

Craig Venter was the first person to have his genome sequenced, that was in 2007. Now, just seven years later, costs have fallen by a factor of 10,000.  Personal genome sequencing is going to become routine regardless of the FDA.

Jonathan June 14, 2014 at 11:10 am

Kha-a-a-a-a-a-a-a-a-a-a-a-a-n!!!

Ray Lopez June 14, 2014 at 12:02 pm
shivuintheshowers June 14, 2014 at 2:57 pm

geeks – genetically genghis is the man

Z June 14, 2014 at 11:34 am

I may live long enough to see DNA results included in applicants CV.

Willitts June 14, 2014 at 1:17 pm

This is how Gattaca started.

Interesting question, though: wouldn’t genetic testing make health insurance premiums more efficient?

Kabal June 14, 2014 at 1:47 pm

Yes — efficient in the sense that the EV between premiums and pay-out risk would be closer to zero for an individual, like a CDS.

However, in the eyes of some, that would be a bug, not a feature.

Some don’t like the idea that people should pay health insurance premiums commensurate to their risk, and find odious the usage of actuarial data in determining premiums (much less how they would feel about genetic data…).

Some even howl about the consideration that young men, who tend to incur lesser health expenses, could/should pay smaller premiums than young women.

RR June 14, 2014 at 2:32 pm

Genetic testing would not make health insurance premiums more efficient in the economic sense. If there is a Pareto-improvement from risk-sharing between risk-tolerant insurance companies and risk-averse households, revealing the true state of nature makes both parties ex-ante worse off since it rules out risk-sharing . Both parties would in fact be willing to pay a non-zero sum ex ante to not have the genetic information become known publicly (as long as they can be assured that neither party can see the information).

Kabal June 14, 2014 at 3:01 pm

Thanks for your reply. This is akin to the stylized model I had in mind:

Person A and Person B are both insured against Disease D by Company K.

Genetic tests reveal Person A has a 5% statistical likelihood of developing this disease in time T, but Person B a 55% likelihood.

In at the absence of genetic tests, Company K would slap them both with the same premium. But now Company K can offer Person A a lower premium, and Person B a higher one. Person A would no longer be subsidizing Person B.

Company K would still be bearing the risk as the risk tolerant party with respect to Person A or Person B, the risk averse parties — the volatility of cash flows over this time horizon T is some large positive number for Company K with either Person A or Person B as the counter-party, but near-zero for Person A or Person B. If disease doesn’t occur over time T, just the premium exchanges hands. Disease occurs, Company K covers. Company K is essentially writing a put option.

Willitts June 15, 2014 at 12:06 am

Wouldnt there be a pooling and a separating equilibrium? Can you say unambiguously that the pooling equilibrium is pareto superior?

In my mind, charging people exactly according to their risk maximizes an objective SWF.

Dismalist June 15, 2014 at 10:00 am

With and without publication of genetic tests we’d get two different Pareto Optima that are distributionally different. The well endowed would be richer and the poorly endowed would be poorer. It’s wrong to think of insurance as redistribution or involving a subsidy, by the way: Ex ante there is none.

Nathan W June 15, 2014 at 2:05 pm

Yes, and that’s why we need laws against that, to prevent discrimination.

lump1 June 15, 2014 at 12:38 pm

Isn’t the whole point of insurance to even out the fickle outcomes of luck? What genes you end up with are one clear product of luck, are they not?

It’s fun to think about a world that had accurate information from the future about the exact fate of every person. Yet even such a world should have an insurance pool that’s decided in a future-blind way, to make life better for the people headed for bad luck, at a cost subsidized largely by those that will be lucky. Or do you not agree?

DIYwut June 14, 2014 at 11:43 am

“Promethease is part of an emerging do-it-yourself”

And I was immediately let down when I couldn’t download the source code :/
I guess I have to wait a couple more years…

sunSunSunGoAway June 14, 2014 at 1:52 pm

Promethease just aggregates info from SNPedia, or so they say. It only costs $5, so you can put in some data, see the results, and probably pretty easily recreate them.

However it’s super uninformative, as far as I can tell. I did not expect to be so disappointed. Adding up the statistically significant additive effects of single SNPs appears to be a pretty terrible way of predicting most quantitative traits, because of a multiple testing problem (many of tiny effects, which are mostly individually undetectable.) Put your results in there and it’s basically meaningless. Promethease doesn’t even try to combine the individual effects. (Combining odds ratios in an additive model…does that require adding the logs of the probabilities or something?)

Look at http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3232052/ for example (lost heritability). Adding individual effects of known SNPs gives you only 5% of the variance on height, whereas it’s generally understood to be about 80% heritable in well-nourished environments. I don’t understand the cited article, but they manage to explain about 40% with a better way of using SNPs (not testing them one-by-one for association with the trait). Maybe some dimensionality reduction technique like PCA regression would work similarly.

I may be wrong.

sunSunSunGoAway June 14, 2014 at 2:09 pm

sorry: penalized regression probably makes more sense than PCA regression, because PCA finds the low dim subspace without looking at the response variable, right?

Also I indiscriminately mixed up binary traits (odds ratios) with continuous traits (expected additive contributions). But what you mostly see on Promethease/ (SNPedia?) is odds ratios.

Anyway, let me know if you find better tools.

ThomasH June 14, 2014 at 11:43 am

I’d love to see the cost benefit analysis the FDA did to decide to disallow DNA testing.

ummm June 14, 2014 at 12:13 pm

They (the left) can’t handle the truth that some people are better than others. Better to be ignorant than concede to reality.

Willitts June 14, 2014 at 1:20 pm

True, but it never occurs to them that the fact we are different can justify greater expenditures on those who are genetically disadvantaged. We’ve come a long way from throwing defective infants off cliffs.

mulp June 14, 2014 at 6:39 pm

Being born disabled is genetic, but being born poor is a choice, so Medicaid must be denied to the poor, but Medicare provided to the disabled?

Willitts June 15, 2014 at 12:13 am

Transfer payments have no disincentive effect on the disabled as they do with the poor.

I’m not necessarily against Medicaid, just highly wary of it. Poverty can certainly reach a point of inability. It’s the margins that concern me. I also like to consider private alternatives.

Nathan W June 15, 2014 at 2:10 pm

Being born poor is a choice? Please explain.

F. Lynx Pardinus June 14, 2014 at 5:50 pm

Ideological Turing test violation detected.

Nathan W June 15, 2014 at 2:09 pm

It has to do with this foundational belief that, even though outcomes are different, that we all have similar inherent value as human beings.

Assuming that, from the perspective of the individual who is born and who cannot themselves control their ancestry, it is distasteful to think that we would further stack the deck against them by allowing perfect market discrimination by insurers.

andrew' June 14, 2014 at 1:39 pm

Simpler: fda will scope creep until they are stopped.

dan1111 June 14, 2014 at 2:27 pm

+1

mulp June 14, 2014 at 5:43 pm

FDA did not disallow DNA sequencing, but they have required evidence that the interpretation is backed by sound evidence.

The brca genes are clearly linked by strong statistical evidence to several cancers of the breast and other organs. Normal genes produce brca hormones that suppress tumors, so the statistical evidence are backed by the evidence of the function of the genes.

But the 23andMe tests were basically making statistical inference on anecdotal data. People who had the tests self reported eye color, etc, and from that, genes were identified as linked to certain traits. A casual relationship but not a causal relationship.

A 50 year experiment with foxes in Russia provides so insight. The “friendly” foxes were allowed to breed while the fearful foxes were skinned for fox fur. In 5 generations, the foxes became friendly, and now 50 years later the foxes are really friendly. But lots of things have changed. Besides fear, also ear floppiness, bone density, jaw size is smaller, fur color, bladder control, strength. A geneticist believe the breeding selection removed an aggressive propagation trait of blast cells which spread throughout the fetus and control the growth of the stem cell line that produces the skin, nails, teeth, bones, and adrenal gland.

Thus one gene variation is linked to skin color, teeth size, jaw size, bone size, and fear.

The 23andMe data might have linked fear to a gene at generation 5, but failed to link it to lack of bladder control until generation 25. Even after 50 years, what exactly is happening in the fox breeding method isn’t fully understood, but its clearly changing the genetic profile of these foxes.

Given you have the same genes as the fox, if 23andMe tells you that the lack of a gene in the embryo will produce a friendlier and less fearful child, would you be happy if the child were also a frail bedwetter?

andrew' June 15, 2014 at 6:15 am

I just want to be able to buy a genetic test.

Nathan W June 15, 2014 at 2:13 pm

I think this experiment is an important early piece of evidence in understand that a small number of genes can affect transcription of numerous other genes (thereby acting as “supergenes” of sorts), if I remember correctly. It precedes the molecular studies which appear to provide strong confirmation of these theories, but I think it made the need for further study quite apparent.

This also tells us that information about a very small number of genes may tell us lots, but it will be very hard to figure out which genes those are.

Douglas Knight June 18, 2014 at 12:26 am

FDA allows myriad to do brca testing, but does not allow 23andme to do brca testing (both Ashkenazi frameshift and SNP). BRCA is exactly the example the FDA gave of what it objects to.

Fully broken (nonsense or framshift) BRCAs are well-established and have large effects, but FDA also allows Myriad to talk about tiny relative risks that are probably bullshit. Most of 23andme is similarly bullshit, but it’s backed by the same standards of evidence as Myriad.

Kabal June 14, 2014 at 12:39 pm

Khan is a noted crime-thinker.

Over the years, I have seen him regularly post images that separate humans into different population groups using genetic data, and/or make posts about how various traits appear in different population groups at different frequencies. This is all just scientific racism.

It is no surprise that he shares a surname with a famed eugenicist villain from Star Trek.

We should be arresting this man for his treachery, not praising him for his nefarious ways.

Willitts June 14, 2014 at 1:18 pm

Well said, Big Brother.

lump1 June 15, 2014 at 12:55 pm

If you think he’s doing bad science, you should raise your objections with his methods.

But it sounds like you think that nobody should ever do a certain kind of science, namely a scientific investigation of correlates between genetic data and human behavior. Or, if somebody does do that kind of science, this is allowed only if they discover something you find politically pleasing, and they must censor their data if they discover something else. Am I understanding this right?

Nathan W June 15, 2014 at 2:16 pm

I am not against such research, but I am against having people who do not respect the general agreement that each human being has inherently equal value, for the fact of being human, even though markets may disagree.

If we can find genes which are strongly linked to warlust, for instance, we could endeavour to access proper counselling for those sick individuals who want war.

More likely a history of abuse, poor social conditions, etc., etc. will continue to be inordinately more relevant than genetics in almost all people and for almost all observed traits, if you’re talking about criminality or what have you.

Keith June 14, 2014 at 2:02 pm

Very cool. Good for him.
I had no idea you could get a Ph.D in feline genetics.

Rhodium June 14, 2014 at 6:20 pm

I am sure most people do not realize how much is online for free. Not only US sites like NCBI or the UCSG genome browser, European EMBL, Japanese ones, like KEGG and tons of research groups with their own software they make available, or run automated requests. The problem is to come up with good hypotheses, the data, tutorials and tools are out there through your browser. If you know of a smart kid looking for a science fair project, this is a direction to investigate.

razib June 14, 2014 at 6:23 pm

“Very cool. Good for him. I had no idea you could get a Ph.D in feline genetics.”

that’s my project. my diploma will say ‘genetics & genomics.’ i consider myself a population genomicist.

Keith June 15, 2014 at 2:09 am

Ah that makes sense now. The Technology Review article mixed that up.

I used to read your blog back in the day when it was 50% genome wide association studies and 50% musings on beautiful women. Good times, good times.

Tom June 15, 2014 at 4:08 am

This could soon turn creepy though. People trying to match up particular sperms with particular eggs in a tube. Or screening and aborting because they’re convinced the baby inherited the wrong genes. Don’t be surprised if a lot of half-baked pseudo-science gets thrown into the mix and the perfect progeny turn out defective. But I suppose they’ll be numerically overwhelmed by the great unscreened anyway.

Marie June 15, 2014 at 11:55 am

Already there on several counts.
As for screening being an elite thing, I believe at least in California the law says you must offer women the fetal screening available now.
I was told 13 years ago by my midwife that I could turn down the test, but she had to offer it.

Tom June 16, 2014 at 2:37 am

Sure, but that’s a very limited screening, for a few genetic sequences that are scientifically proven to be severely debilitating, and globally only a small portion of embryos are getting even that much screening. I’m all for that kind of screening.

The potential creepiness that I’m suggesting is that people could do their own screening of embryos or sperms and eggs to try to ensure their offspring receive whatever genes they consider to be better for whatever reason.

lump1 June 15, 2014 at 1:02 pm

Yeah, I think you’re right, but embryo selection – even if guided by somewhat shoddy heuristics – doesn’t seem to me to pose a danger. In typical pregnancies, a selection process happens anyway, and I think we know enough about genetics to at least steer the selection away form the worst of the possible genetic illnesses, without introducing extra risk. If I thought that the selection would *increase* the likelihood of a “defective” (your word) offspring, then of course I would be far less blase about the whole thing.

Tom June 16, 2014 at 2:48 am

I somewhat exaggerated for effect, but I think it’s possible that if parents screen sperms and eggs to get what they consider to be the best possible selection from among their genes, they could unknowingly pick problematic genes that might have been weeded out in a natural process.

jon livesey June 16, 2014 at 4:31 pm

“Craig Venter was the first person to have his genome sequenced, that was in 2007. Now, just seven years later, costs have fallen by a factor of 10,000.”

Now imagine what this has done to the data storage industry.

Comments on this entry are closed.

Previous post:

Next post: