This is a bit difficult, but it shows how important diversity is in the subjects science uses in its experiments
The colours in the graphs are difference populations, AA, African American; EA, European American. X axis is how genetically similar they are to the "average" of 35,000 individuals in a diverse biobank (ATLAS) at the University of California, Los Angeles. The Y axis shows how good we are at interpreting the effect of their genome. In this case it is just height, but the same effect is to be expected in any medical multilocus genetic test, and quite possibly a lot of single locus tests as well.
Examining the accuracy of a genetic score predicting height in individuals with different ancestries. Polygenic scores (PGSs) are estimates of the cumulative effect of thousands of genetic variants on the likelihood of having a certain trait. A PGS predicting height was trained on a subset of people annotated as white British in the UK Biobank. PGS accuracy was then tested for individuals (dots) with different ancestries in the ATLAS biobank in Los Angeles, California.
a, PGS accuracy decreased with genetic distance (genetic dissimilarity;
di) from the training population, in clusters with homogeneous or admixed and unclassified genetic ancestries.
b, Individual PGS accuracy decreases with genetic distance from the training population across the entire ATLAS biobank.
R and
P refer to the correlation between genetic distance and individual-level PGS accuracy and its statistical significance, respectively. AA, African American; EA, European American; EAA, East Asian American; GIA, genetically inferred ancestry; HL, Hispanic or Latino American; SAA, South Asian American.
Source