Genes, ancestry and metabolism: why not everyone needs the same diet

Two colleagues sit in the same office kitchen in the morning. Both eat the same thing: a roll with jam and a coffee. One works with full focus until lunch. The other sits there at half past ten and feels as if someone had pulled the plug.

Both are healthy. Both slept about the same. Both ate the same thing.

You probably know this scene. Maybe from your partner, who can eat pasta like a marathon runner while one plate of it sends you into an afternoon nap. And at some point comes the question I am asked often in my practice: is this about where I come from?

The answer has two halves. First: yes, evolution has adapted metabolism to very different menus, and we can read that in the genome. Second: no, so far nothing follows from it that gives you a personal eating plan. Both halves belong together.

The same meal, two very different curves

Nutrition counselling carries a quiet assumption: a food has a value, and that value holds for everyone. The glycaemic index sits in tables as if it were a property of the apple. But it is also a property of the person eating it.

Since glucose sensors exist, this can be measured. The numbers are more striking than I had expected.

That leaves the question of where this spread comes from. Part of it does sit in the genome.

Starch: the amylase gene as an adaptation to the plate

Chew a piece of white bread long enough without swallowing. It turns sweet. That is chemistry: your saliva carries amylase, an enzyme that breaks long starch chains into short sugar units. Digestion begins in the mouth.

What is special about the gene behind it: it does not come in two copies as usual. In some people there are one or two, in others more than ten. As in a supermarket, with the same crowd it makes a difference whether two or twelve checkouts are open.

And now comes the part that test providers rarely mention. In 2014 a paper appeared reporting that a low AMY1 copy number goes along with a higher obesity risk, calculated as roughly an eightfold difference between the top and the bottom tenth. That was the birth of many reports about the carbohydrate type.

A year later another group measured the same locus again with better technology. With sufficient statistical power the link with body weight could not be confirmed. A Mexican paper in 597 children found the original pattern again, but it remained an association study.

What remains is the adaptation story, and it is strong: starch digestion begins in the mouth, and how vigorously it begins there has a long prehistory.

Milk: an adaptation that arose independently more than once

Most people can no longer digest milk sugar as adults. That turns our sense of normal upside down. In mammals, switching off lactase production after weaning is the default. The exception is to keep producing this enzyme for a lifetime.

Picture a light switch that goes off in almost all mammals after weaning. In some humans it is jammed and stays on. Not because they are healthier, but because their ancestors had milk available.

How strong this signal is shows in a paper by Tal Bersaglieri. In samples of northern European descent the associated alleles mark a haplotype at about 77 percent that runs almost uninterrupted over more than one million bases. The authors call it one of the strongest selection signals in the human genome, arising within the last 5,000 to 10,000 years.

And then the neat story became complicated. In 2022 Richard Evershed and a large team analysed around 7,000 fat residues from pottery at more than 550 sites to map milk use across 9,000 years.

The result did not match the expectation. A selection model following the intensity of milk use explained the allele frequencies no better than a model with steady selection. In the UK Biobank the genotype was only weakly linked to milk consumption. As a hypothesis the authors propose periods of famine and disease burden. A review by Laure Ségurel and Céline Bon names further open contradictions, such as low frequencies among Central Asian herding peoples.

That makes it clear why tolerance here is the exception and not the rule.

The Arctic: when carbohydrates become metabolically expensive

If there is a documented case in which a population is genetically set up for a very fat rich diet, it is in the Arctic. And this very case shows most clearly why no recommendation follows from it.

A third finding fits with this. Matteo Fumagalli and colleagues searched the genomes of Greenlandic Inuit for selection signals. The strongest lay at the fatty acid desaturases, the enzymes for polyunsaturated fatty acids. The selected alleles were connected to weight and body height.

And now you see why the strongest finding in this article of all things yields no recipe.

Omega-3: why some people can build more EPA themselves

You may have heard that linseed oil is not enough as an omega-3 source because the body converts the plant precursor into EPA and DHA only poorly. That is true in principle. What is rarely added: how poorly this conversion runs differs from person to person.

Two neighbouring genes are responsible, FADS1 and FADS2. They encode the enzymes that form the bottleneck of this conversion chain. Picture a workshop: the number and the speed of the machines decide the output.

How different the frequencies are was studied by Kumar Kothapalli. A regulatory polymorphism in FADS2 that favours endogenous conversion was present at 68 percent in a largely vegetarian Indian sample compared with 18 percent in a US sample. Calculated globally: around 70 percent in South Asian, 53 in African, 29 in East Asian and 17 percent in European samples.

And the direction of selection was not the same everywhere. Michael Buckley compared FADS data from present-day and Bronze Age Europeans: in Europe other alleles were selected than in South Asia or Greenland, with the opposite effect on the fatty acid profile. A paper by Iain Mathieson on ancient DNA from 230 West Eurasians shows in general that diet related gene loci are among the most striking targets of selection.

Something modest but useful follows from this in practice. If you eat vegetarian or vegan, do not rely on your own conversion being enough. Whether it is enough is not visible in your family tree but in the omega-3 index in your blood. The physiology is in the article ALA, EPA and DHA, the question of sources in the article Omega-3 from plants and animals.

And with that you know why two people with the same intake do not have to end up with the same blood value.

Alcohol, folate, cholesterol: three variants, three levels of evidence

Up to here this was about adaptation across millennia. Now it gets more concrete. Some gene variants do play a role in everyday life, but the evidence behind them differs a lot in strength.

And with that you have a grid for every future gene variant headline.

Why ancestry is still not a set of dietary instructions

Now comes the section I consider the most important. This is where the discussion regularly slips, either into determinism or into a complete shrug.

The first reason is statistical. Noah Rosenberg and colleagues studied 1,056 people from 52 populations at 377 gene loci. 93 to 95 percent of genetic variation lies between individuals within a population, only 3 to 5 percent between the large geographic groups. Two people from the same region can be genetically further apart than two people from different continents.

The second reason held me back the most. The best known ancestry and metabolism pattern is the increased diabetes risk among South Asian migrants: illness typically five to ten years earlier and at a lower BMI.

In 2004 the WHO therefore made a remarkably cautious decision. An expert consultation examined separate BMI cut-offs for Asian populations. The threshold for observed increased risk lay between 22 and 25 depending on the population. Instead of a new number for everyone, additional points for action were proposed at 23.0, 27.5, 32.5 and 37.5.

A meta-analysis by Keiichi Kodama across 74 cohorts with 3,813 people fits with this: insulin sensitivity and insulin response sit at different points on the same curve in different groups. The authors suspect a genetic background but did not measure any genetic markers.

And finally the part of metabolism that has the most direct connection to eating. Daphna Rothschild analysed data from 1,046 people with very different ancestries living in a shared environment. The gut microbiome was not significantly linked to ancestry. In contrast, the microbiomes of unrelated people sharing a household resembled each other.

And that is why the next question is the right one: if ancestry reveals nothing, can a gene test at least do it?

What DNA diet tests promise and what has been examined

On the providers' websites it sounds convincing. You spit into a tube and after a few weeks you get a classification: fat burner, carbohydrate burner or mixed type, plus a nutrient distribution in percent. The price is usually between 150 and 300 euros.

I searched specifically for the primary literature that defines and examines this three way split. I did not find it. As far as I can trace it, the categories come from marketing and not from the specialist literature. That does not automatically mean they are wrong. It means that nobody has tested them robustly so far.

What has been examined is the core claim behind them: that the genotype predicts which diet works better for you.

This criticism does not come from outside. A group of authors from nutrigenetics itself stated in 2017 that the field is unregulated and that beyond commercial codes of practice no defined standards exist. A review by Dong Wang and Frank Hu names a lack of reproducibility and the still missing proof of additional benefit compared with classical nutrition intervention.

That leaves the question of what works instead.

Measure instead of guess: the path that really is individual

If you have read this far, a slightly frustrating feeling may have set in. The differences are real, the genes barely explain them, and your family tree explains them even less. What is left?

Quite a lot. Just not from the laboratory of the past, but from your own body.

And then the number that sums up the article for me. In PREDICT 1 genetic variants contributed 9.5 percent to predicting the glucose response, and 0.8 percent for the blood lipids. Person related factors such as the gut microbiome explained more variance there than the macronutrients of the meal itself.

Put differently: the sensor on your arm can show you more about your metabolism after two weeks than a saliva test.

If you would rather not do this on your own: below this article there is the option to arrange an appointment.

And now you know why the most honest answer to the question about the right diet is not in your family tree, but in a curve you can produce yourself.

Common questions about genes, ancestry and nutrition

Where this topic can connect

The question of the individual metabolic response comes up again in several areas: with insulin action, with blood sugar, with fatty acids and with training. From here several paths lead onwards.