Mapping the 'Mexican genome': the questions a 2009 launch skipped

On May 11, 2009, Mexico's National Institute of Genomic Medicine, INMEGEN, presented what it called the map of the genome of Mexicans. President Felipe Calderón attended. The country was still in the middle of the H1N1 flu emergency that had begun in April, and the event was pitched as proof that Mexico could do frontier science for its own people. The same day, the paper behind it went online in the Proceedings of the National Academy of Sciences (doi:10.1073/pnas.0903045106).

Not everyone applauded. Within two weeks a column in the Mexican daily La Jornada, written by a researcher at the technology watchdog ETC Group, called the whole thing a farce. Her case, roughly: the headline finding, that Mexicans descend from Indigenous, European and some African ancestors, was hardly news after five centuries; the gene-hunting methods it would feed had a poor record of predicting disease; the tools for turning results into products were patented by a few foreign firms; and nobody at the podium had said who would benefit.

That was a polemic. Read sixteen years later, it is also a fair list of the questions genomics in Mexico, and in much of the global South, has had to answer since. What follows is what the study actually did, the law meant to protect it, and five questions the launch skipped.

What was actually mapped

"Map" oversold it. The Mexican Genome Diversity Project, run by an institute founded only in 2004, did not sequence anyone's genome. It took blood from 300 unrelated people who identified as mestizo, in six states chosen to span the country (Sonora and Zacatecas in the north, Guanajuato in the center, Guerrero, Veracruz and Yucatán further south), plus 30 Zapotec people from Oaxaca as an Indigenous reference group. Each sample went on an Affymetrix chip that reads about 100,000 preselected single-letter variants, or SNPs. A human genome has around three billion letters.

That is enough for useful population genetics, and the paper did some. On average the mestizo samples came out about 55% Indigenous American and 42% European in ancestry, with small African and East Asian fractions. The averages hid real regional spread: Indigenous ancestry ran to about two-thirds in the Guerrero sample and just over a third in Sonora. The authors also found 89 common variants in the mestizo groups that were absent from the international HapMap reference populations. Every variant unique to the Zapotec sample turned up among the mestizos too, which tells you where the distinctively "Mexican" variants came from.

The practical argument was about cost. Genome-wide association studies don't read every variant. They read "tag" SNPs that stand in for neighbors usually inherited along with them. The paper found that tags chosen from all the HapMap populations combined would already capture about 96% of common variation in Mexicans, but only by genotyping more markers than necessary. A Mexican reference, the authors argued, would give the same coverage with fewer tags. Sensible, and a long way from the press-conference talk of medicines designed for Mexicans.

Genomic sovereignty, as written in 2008

The year before, Congress had added Article 317 Bis to the General Health Law, published in July 2008. Sending human tissue out of Mexico for population genomic studies now needed approval from a Mexican research institution and a permit from the health authorities, and the health ministry was to keep a registry of permits, with INMEGEN as its federal adviser and national reference center. The stated background was concern about foreign teams collecting blood in Mexico, including from Indigenous communities, without Mexican approval.

The idea was called genomic sovereignty, and its limits were visible from the start.

• It covers material, not information. The PNAS paper itself announced that its genotypes and raw data files could be downloaded from the institute's FTP server. A permit regime for test tubes does nothing once the data are online.
• It makes the state the rights holder. Individuals and Indigenous peoples do not appear as parties with a say of their own.
• It is silent on patents and on any sharing of benefits.
• It stops at the border. The third phase of HapMap had already added samples from people of Mexican ancestry living in Los Angeles, and tens of millions of people of Mexican descent live in the United States.

Five questions the launch skipped

1. Better medicine for whom?

The diseases genomic medicine promised to tackle in Mexico, diabetes and obesity above all, are driven heavily by what people eat and drink, what they earn and where they live. In April 2009 a cluster of commentaries in the New England Journal of Medicine had asked how much the hits from association studies actually explained. For most common diseases the honest answer was: not much per variant.

Mexican genetics did produce one striking result. In a 2014 Nature paper, the SIGMA consortium, which paired the Broad Institute in Massachusetts with Mexican institutes, reported variants in a gene called SLC16A11 that raise the risk of type 2 diabetes. The risk version shows up in about half of Native American samples and is rare in European and African ones, which is why years of European-dominated studies had missed it. It entered the human lineage, of all routes, through interbreeding with Neanderthals. It is a real discovery and, so far, a research lead rather than a treatment. The measures that have actually gone after Mexico's diabetes problem at population scale were a tax on sugary drinks in 2014 and black warning labels on packaged food from 2020.

2. Who consented, and to what?

The Zapotec participants were there as a comparison group, yet their DNA did much of the explanatory work: the variants that made the Mexican samples distinctive were Indigenous variants. The paper covers community engagement in a single sentence, noting that an "ad hoc" process of consultation was carried out. Critics at the time pointed out that the institute's own account of whether Indigenous people were being sampled had shifted over the life of the project.

Consent in population genetics is not paperwork. In 2010 Arizona State University settled with members of the Havasupai Tribe, who had given blood for diabetes research and later found it used to study schizophrenia, inbreeding and migration; the samples were returned. The wider pattern, research designed by a well-equipped institution and carried out on people with little power to question it, is what made a US company's trial of GM-rice proteins on sick children in Peru so contentious.

3. Who owns the results?

In 2009 the chips, sequencers and software that genomic medicine depended on came from a handful of mostly American companies, and the quickest commercial outlet for association findings was the genetic test. Since 2013, the US Supreme Court's Myriad decision has ruled out patents on naturally occurring DNA sequences as such. It left plenty else open: methods, diagnostics, drugs and synthetic sequences, as the story of the first synthetic cell shows. A variant common in Mexicans cannot be patented in the United States. A test or a drug built on it can be, by whoever gets there first, and nothing in Article 317 Bis says that has to be anyone in Mexico.

4. Where does the data go?

Two papers published side by side in Nature in October 2023 show how far the scale has moved, and in which directions.

The first came from the Mexico City Prospective Study, a cohort of more than 150,000 adults recruited around the turn of the century in the boroughs of Coyoacán and Iztapalapa and run with the University of Oxford. Every participant was genotyped and exome-sequenced, and nearly 10,000 had their whole genomes sequenced, with the Regeneron Genetics Center, the research arm of the US drugmaker, as a major partner. The competing-interest statement lists authors employed by Regeneron, AstraZeneca and AbbVie. Ancestry-specific allele frequencies are free to browse online, and the authors present the resource as useful both in Mexico and in the United States, where most Hispanic people are of Mexican descent.

The second was the Mexican Biobank: 6,057 people from 898 rural and urban localities in all 32 states, genotyped at 1.8 million markers, linked to health data and led by Mexican geneticists. For several of the 22 traits it studied, predictions built on its own association results did better than ones borrowed from the UK Biobank.

Both are serious science, and neither fits the logic of the 2008 law. What travels now is data and access agreements, not tubes.

5. Who shares in the benefits?

Here is the twist. The international rules on benefit-sharing for genetic resources, the Convention on Biological Diversity and its Nagoya Protocol, leave human genetic material out of scope. So the Cali Fund agreed at COP16 in late 2024, which asks large companies that profit from digital sequence information to pay 1% of profits or 0.1% of revenue, with at least half earmarked for Indigenous peoples and local communities, covers sequences from microbes, plants and animals, not people. The complaint behind it is the one made in 2009: sequences flow in free, products flow out under patent. Human genomes are simply outside the deal.

The most useful thinking on the human side has come from Indigenous researchers. The CARE principles for Indigenous data governance, published in 2019 by the Global Indigenous Data Alliance (collective benefit, authority to control, responsibility, ethics), treat data about a people as something that people should have a say over. It is a different idea of sovereignty from Mexico's in 2008: held by communities rather than the state, and attached to data rather than samples. The same demand for community consent runs through the argument over releasing gene drives.

What the critics got right, and what they didn't

The 2009 critique underrated one thing. Studying Mexican genomes did turn up biology that studies of Europeans had missed, and local reference data do make genetic predictions work better for people in Mexico. "Mexico needs its own data" turned out to be scientifically true.

What it got right was the question nobody at the launch answered. A genome map is infrastructure. Who uses it, who profits from it, and whether the people whose blood built it see anything back get settled later, in data-access agreements, licensing deals and patent filings that never get a press conference. Sixteen years on, Mexico has far better maps. Who they are for is still an open question.