It's an uncommonly bright day in Pacifica, California, the often fog-bound beach town just south of San Francisco where tech guru and Wired magazine maverick editor Kevin Kelly is showing me his DNA. And I'm showing him mine.

Welcome to the new world of genetic voyeurism, where those of us who are DNA curious can now peek inside another person's double helix and glimpse his or her DNA.

You 2.0 Series Comparison Shopping for Your Future Part 1: Personal genetic tests are proliferating; some are even online. Do they really tell you anything? I'm Doomed. Or Not. Part 2: Commercial genetic tests tell me I have high, medium, and low risk of heart attack. What gives?

Kevin and I are perusing our results on 23andMe and DeCodeMe, online genetic-testing sites that have analyzed our DNA and posted our results on password-protected Web pages. We check out risk factors for everything from heart attack to diabetes, comparing our disease results, me using my laptop and Kevin looking at two giant flat-screen monitors.

Kevin, who writes the The Quantified Self blog, is typically gung ho about this new technology and information but admits to being underwhelmed so far. "I'm not learning that much about myself," he says, but he believes this will change. "It's like the first personal computers or fax machines."

I was able to get these tests gratis as a journalist or paid for by my publication, but Kevin is an actual customer. He paid nearly $1,000 each to 23andMe and DeCodeMe. "This is too expensive for what you get," he tells me, though he points out that the first PCs cost $4,000 or $5,000.

Kevin and I are both relatively clear of scary secrets in terms of diseases hidden in our genes, though if you read my last column, you'll see that I might have a high risk of heart attack—or not, depending on which testing company one is to believe.

I'll let Kevin reveal what he wants to about his results for disease; what I'd like to share right now is how we fared on those genetic variations that might be called recreational or just plain bizarre. These include gene markers for wet or dry earwax, the sprinter's gene, and a marker that increases one's chance of becoming a heroin addict.

I'm not kidding. Skip ahead to the chart of results below, and you'll see that both Kevin and I are at high risk of drug addiction, according to a 2004 study conducted in Sweden. Here is what 23andMe says about this test:

In the brain, heroin is converted to morphine, an opioid painkiller. Morphine acts by signaling through a receptor encoded by the gene OPRM1. Different versions of the OPRM1 gene are thought to affect how much morphine one needs to feel a given effect. This study of 139 heroin addicts (primarily Swedes) and 170 nonaddicts found that people with one or two copies of the G version of the SNP rs1799971 have almost 2.9 times the odds of being a heroin addict.

Here is how Kevin's results are presented on his 23andMe site:



A 2.9 times normal risk factor is relatively high, though even 23andMe gives the validity of this data only two stars (out of four) because the researchers tested only about 300 people. That gives these results a very low statistical power to predict whether others, such as Kevin and myself, will really become addicts. 23andMe calls such findings preliminary, which makes one wonder why such a test is even included. 23andMe offers the following caveat:

Preliminary Research This is a Preliminary Research topic and includes results of studies that still need to be confirmed by the scientific community. It also includes topics where there may be contradictory evidence. The results of these studies are not conclusive.

Just for the record, Kevin and I are not addicted to heroin.

Most of these "fun" or oddball traits appear only on 23andMe, which features results for 60 genetic traits, compared with 26 for DeCodeMe and 17 for Navigenics (these numbers vary slightly according to sex).

23andMe has close ties to Google and tends to treat nearly all genetic results as equal—that is, it lists and describes earwax alongside age-related macular degeneration and colorectal cancer. Each of 60 traits it offers is accompanied by detailed information about the trait, studies, risk factors, and links to more information.

At the opposite end of the spectrum is Navigenics, which offers only information on medical conditions—no bitter taste or heroin addiction. In between the two is DeCodeMe, which offers mostly disease information but throws in some recreational results—bitter taste, for instance.

23andMe and DeCodeMe also provide results for ancestral data that traces a customer's genetic lineage back to migration patterns of early humans coming out of Africa 50,000 years ago and spreading out to Asia, the Americas, Oceania, and Europe.

DeCodeMe tracked my DNA to the most common markers for Western Europeans, with the closest ethnic association through my father (my Y chromosome) connecting me to Scotland—where my ancestors do come from.

Through my mother (my mitochondrial DNA), I am also associated with a band of humans that emerged in Europe 15,000 years ago. Luminaries from this haplogroup—a cluster of people sharing similar ancestral patterns of DNA—include Saint Luke, Marie Antoinette, and Warren Buffett.

Kevin, who is of Irish ancestry, belongs to the same "tribe" as I do through our fathers, but his mother's lineage links him with a group that also includes Kurds, Druze, and Ashkenazic Jews.

For many other attributes, Kevin and I tested the same, for better or for worse. Both of us have a high probability of having blue eyes and of being lactose tolerant, things we knew even without a genetic test.

---

DNA Traits Comparison: David Ewing Duncan and Kevin Kelly

Trait	Gene Marker	Risk	D.E.D.	K.K.	Risk Factor
Alcohol flush*	rs671	A	GG	GG	Normal/No flush
Avoiding errors	rs1800497	A	GG	GG	Learns to avoid errors
Blue eyes*	rs12913832	G	GG	GG	Blue eyes
Bitter taste*	rs713598	C	CC	CG	No bitter taste/ Bitter taste
Caffeine metabolizer	rs762551	C	AA	AC	Fast/Slow
Earwax	rs17822931	C	CC	CC	Wet
Heroin addiction	rs1799971	G	AG	AG	Substantially higher risk
Endurance (or Sprint)	rs1815739	T	TT	CT	High endurance/ Sprinter
Intelligence	rs363050	G	GG	GG	Lower I.Q. (3 points)
Lactose intolerance*	rs4988235	G	AG	AA	Lactose tolerant
Novelty Seeking**	rs6280	T	CT	TT	Med. Risk-Taking/High Risk

^{All traits are from 23andMe.com, except:
*Also on DeCodeMe.com
**Not on either site}

---

Also, neither of us have a variation that causes some people's cheeks to become flushed when they drink alcohol, and both of us have a proclivity to learn from our mistakes—a useful thing for writers and geneticists.

Less attractive is a DNA marker linking us to a reduction of three points in our I.Q. scores, whatever that means.

We diverge, however, on a marker for risk taking—Kevin is a bigger risk-taker. He also has a gene marker often found in sprinters, while I have the version of the same marker associated with athletic endurance.

Then there is one of my favorites—a marker associated with rapid caffeine metabolism (as I suck down another latte). I have a genetic variation linked to being able to drink coffee all day with no added risk of heart attack. Poor Kevin has a variant that links caffeine consumption to an increased risk of heart attack. (If heroin doesn't get him, a double espresso will.)

About half of these traits get 23andMe's top 4-star rating, though even this rating includes studies with only 1,000 people tested to make a link between a genetic marker and a trait or a disease. Most geneticists consider a test group of only 1,000 people to have a low statistical strength compared with studies in which thousands or tens of thousands of subjects were tested.

Analyzing too few people means that random outliers—those who have or don't have a gene or a disease—can overly influence the results by causing the risk factors to be too low or too high. Imagine polling 100 people in Harlingen, Texas, about their choice for president: You would get a skewed result compared with a poll in a larger or more statistically relevant population.

23andMe's rating criteria:

Established Research These topics meet our criteria for findings that are very likely to reflect real effects. 4 stars: At least two studies that examined more than 1,000 people with the trait/condition, or smaller studies where there is a consensus that the effect is real. Preliminary Research Includes results of studies that still need to be confirmed by the scientific community. 3 stars: More than 1,000 people with the trait/condition were studied. However, the effect has not yet been confirmed in a second independent study of similar size. 2 stars: Fewer than 1,000 people with the trait/condition were studied. 1 star: Fewer than 100 people with the trait/condition were studied.

"I am interested in finding out how to quantify who I am as an individual," Kevin tells me. "That's why I'm doing this. It's not useful if this information is incomplete or inaccurate, but the best way to fix this is with more and better information."

None of the sites are saying how many people are paying $1,000 (the price that 23andMe and DeCodeMe charge) or $2,500 (Navigenics' price) and getting their results, though I suspect it's not a huge number. Some people argue that genetics now is where cell phones were in the mid-1980s: Remember how rare it was then to see someone using one of the first clunky mobile phones? Then cell phones got smaller, sleeker, and more useful, and now they're practically ubiquitous.

One's genes, though, are not cell phones or computers. They are part of what makes us who we are and can give us clues to how we will live and die. That's why it's crucial to get this right.

---

Next week (the fourth and final part of this series): Why the online genetic testing companies are important.