European Protein Genetic Scores Perform About 20 Percent Better in Europeans Than in Arabs

If you searched 'do European polygenic risk scores work for Arab populations', the answer from a 2023 Qatar Biobank study is: partly, with a measurable penalty. A protein polygenic score built from European data explained about one-third of the genetic heritability of blood protein levels across the proteins studied and performed roughly 20 percent better in Europeans than in Arabs. The study measured 1,301 proteins in 2,935 samples with whole-genome sequencing and the SOMAscan aptamer platform.

Do European polygenic risk scores work for Arab populations?

The short answer is yes for the broad signal and no for the precision. The researchers took protein quantitative trait loci from European studies, called pQTLs, and checked them in an Arab cohort from Qatar Biobank. Most of the shared signals looked real across populations: 81.8 percent of shared pQTLs colocalized. That means the same stretch of DNA was usually driving the protein level in both groups.

But when the team built a protein polygenic score from European weights and applied it to both populations, it performed about 20 percent better in Europeans. That is the gap between 'this score still contains useful information' and 'this score means the same thing for everyone'. It does not.

As an illustration, a scatter plot of score against measured protein levels would show European points clustering closer to the line and Arab points showing more scatter. A high score would still tend to mean higher protein, but with more exceptions and more overlap between high and low groups.

A genetic score is a rank, not a measurement

The common mistake is treating a polygenic score like a blood glucose value. It is not a portable single number. It is a weighted sum of many small genetic effects. The weights came from a European discovery sample. They were estimated against a specific pattern of linked variants, allele frequencies, and population structure.

When that weighted sum is applied to a different population, three things can change.

• Linkage disequilibrium. A marker SNP in Europeans may not tag the causal variant as tightly in Arab genomes.
• Allele frequency. A variant that gets heavy weight because it is common in Europeans may be rare or absent elsewhere, so it contributes less information.
• Environment and measurement. Diet, body composition, medications, and aptamer binding can affect measured protein levels in ways that are not genetic.

A polygenic score is a rank inside a reference population, not a measurement like a blood glucose value. Change the population and the rank can shift.

This is why the exact percentile matters less than the population the percentile was calculated in. A score at the 90th percentile in a European reference may correspond to a different absolute protein level, or a different disease probability, in an Arab reference.

What the study did and did not show

This paper is about protein levels, not directly about heart attack or diabetes risk. Proteins sit between genetic variants and disease, so they are useful intermediate traits. The researchers used 2,935 samples from Qatar Biobank, measured 1,301 blood-circulating proteins on SOMAscan, and compared protein quantitative trait loci and protein polygenic scores across populations.

The headline numbers:

• 81.8 percent of shared pQTL signals colocalized between European and Arab samples.
• European-derived protein scores explained about one-third of the genetic protein heritability across the proteins studied.
• The same scores performed about 20 percent better in Europeans than in Arabs.

That is encouraging and limiting at the same time. Most of the real biological signal transfers, but the calibration does not fully survive the trip.

Should you ignore a polygenic score if you are Arab?

No, but you should read it with less certainty about the exact rank. The protein scores in this study still explained meaningful heritability in Arabs. They did not become useless. But the drop means a score that looks extremely high in a generic report may be less extreme when re-ranked in the right population.

If you see a polygenic risk score, check whether the company or study says which population the reference distribution came from. If it does not, you are looking at an uncalibrated number. That is especially true for people with mixed ancestry, because even a single population label may not capture which reference is closest to your genome.

A practical rule: a polygenic score should change how much attention a number gets, not replace the number itself. If a protein score is high, the obvious next step is measuring that protein directly. Worth asking a doctor about.

When the simple 20 percent answer does not hold

The 20 percent figure is an average across proteins. Some proteins likely transferred much better, some much worse. The abstract does not report per-protein variation. It also tested one Arab cohort from Qatar. 'Arab' is broad. Populations in the Gulf, North Africa, and the Levant have different genetic histories, and mixed ancestry shifts the result further. A person whose ancestry is close to the discovery population may get a score that is almost as well calibrated. A person with more distant ancestry may get a worse score.

Disease polygenic risk scores can be less portable than protein scores because disease outcomes also involve environment, diagnosis, and healthcare access. The protein layer is closer to the genome, so if the protein score still loses 20 percent, a disease score may lose more. Do not assume the protein number applies unchanged to a disease prediction.