Next time you want to start a bar fight, proclaim to everyone within earshot that “race is not real; it is just a social and cultural construct and has no biological validity.” Then duck before you get punched in the face. . . . but as you're avoiding injury try to hand your would-be assailants a new paper published online this afternoon by the journal Clinical Pharmacology & Therapeutics, which concludes that classifying people by the crude category of race—as in, of African, Asian or European ancestry—for medical purposes, as some people want to do, is really, really stupid.

It would seem that nothing is as obvious as the reality of race. But while differences in skin color and facial features that are characteristic of the continent you trace your ancestry to—Africa, Asia or Europe—are clear, they are also superficial and potentially misleading. As I wrote way back in 1995, where you draw the lines between races depends on which trait you notice. It happens that skin color and facial features are obvious, which is why we draw the dividing lines where we do.

But how you group people depends on which traits you focus on: sorting people according to one set of traits produces different groupings than sorting them by different but equally valid traits. Say you decide that the distinguishing trait is the gene for hemoglobin. If you divide humankind by which of two forms of the gene each person has, then equatorial Africans, Italians and Greeks fall into the “sickle-cell race;” Swedes and South Africa’s Xhosas (Nelson Mandela’s ethnic group) are in the healthy-hemoglobin race. Or how about dividing humanity by who has epicanthic eye folds, which produce the "Asian" eye? Then the !Kung San (Bushmen) belong with the Japanese and Chinese. Or say you sort humanity by the presence of the lactase gene. Then Norwegians, Arabians, north Indians and the Fulani of northern Nigeria are in one race, while everyone else—other Africans, Japanese, Native Americans—forms the no-lactase race. Depending on which trait you choose to demarcate races, “you won't get anything that remotely tracks conventional [race]categories,” anthropologist Alan Goodman told me back then.

The point is, “race” is too broad a category, and three is not enough. There are indeed real genetic, biological differences between people, but at the level of population or ethnicity, not race. As Goodman put it then, “race, as a way of organizing [what we know about that variation], is incredibly simplified . . . There is no organizing principle by which you could put 5 billion people into so few categories in a way that would tell you anything important about humankind’s diversity.”

Which brings us to the new study. Scientists at the J. Craig Venter Institute got the cool idea of analyzing the genomes of two white guys who, according to the conventional racial categories, belong to the same race. The two are Venter himself and James Watson, co-discoverer of the double-helix structure of DNA. Venter led the private effort to sequence the human genome, winding up in a tie with the public project to do the same.

It happens that the genomes of both men are in the public domain. Watson agreed to have his sequenced and published last year, with Venter right behind. So what do the genomes reveal?

The two men metabolize drugs, including antidepressants, codeine, antipsychotics and the cancer drug tamoxifen, differently. Venter has two functional copies of the CYP2D6 form of the cytochrome P-450 gene, which metabolizes more than 75 percent of drugs, while Watson has two copies of the more-sluggish variant of the gene. That’s rare for Caucasians (only 3 percent of whites have the sluggish version), but common in East Asians (49 percent of whom have it). Funny, Watson doesn’t look Chinese. But if Watson’s doctor decided to use race-based medicine to predict how he would metabolize drugs, she’d say, well, we have a white guy here, and whites rarely have the sluggish version, so I’ll assume Watson doesn’t have it either. As a result, the drug would stay in Watson’s system longer, with stronger effects compared to someone in whom the drug was quickly metabolized and cleared from the body. “It is unlikely that a doctor would guess that optimal drug dosages might differ for Drs. Watson and Venter,” the scientists write.

That’s why Venter and colleagues conclude that race is too crude a proxy for what genetic group—ethnicity or, as biologists say, population—someone belongs to. It is imperative to “go beyond simplistic ethnic categorization,” they write, since that can be seriously—and perhaps fatally—misleading. (In the U.S., some 100,000 people a year die of adverse drug reactions, many caused by an inability to properly metabolize the medication because of a particular CYP2D6 variation.) “Race/ethnicity should be considered only a makeshift solution for personalized genomics because it is too approximate,” they write.

We are on the verge of the $1,000 genome, and existing tests for CYP variants and other genes relevant to medicine cost a few hundred dollars. If we want to tailor medical care to someone’s genes, we should make sure what those genes are. And not assume that all white guys have “Caucasian” genes.