How reliable is a region estimate?

A young man in Shah Alam opens his results. His estimate says 62 per cent one region, 30 per cent another, 8 per cent a third. His first thought is: "Is that exact?" His second: "What does a number like that really mean?"

The story

A region estimate compares your DNA with reference groups: sets of people whose DNA is already known and who are said to come from a certain place. The test then says which groups you look most like, and in what mix.

Think of it as a weather forecast, not a photograph. It is a best guess with a range around it. Three things shape how good the guess is.

1. The reference groups. Scientists point out that results depend on which groups are chosen and what they are called (Mathieson and Scally, 2020). If the right group is missing or small, the test may place you with a nearby, better-covered group. This matters for Southeast Asia, where many peoples are under-represented in public and commercial databases. That is one reason it helps to include Malaysians in reference sets, and why we work with Malaysians we have tested.

2. How fine the picture is. A big picture, such as "East Asia" or "South Asia", is much more reliable than a fine one, such as a particular province or village. The same paper notes that at fine scales, we are much less sure that peoples have stayed the same over thousands of years. Migration and mixing have moved people around (Pickrell and Reich, 2014).

3. The method. Different methods use the same DNA in different ways. Fine-scale methods can estimate even the timing of mixing events (Hellenthal and colleagues, 2014), but a number is still a summary.

So a result like "62 per cent" is best read as "mostly this region, with a range". Treat 62 as about "most of it", not as an exact figure. It is a statement about similarity, and not a count of ancestors.

Where you can still see it today

In your results: a region estimate with percentages and, ideally, a range.
In the words your report uses: "closest to", "most similar to", not "you are".
In this site's stories, which tell the history so you can set the number in context.

What DNA can tell, and what it cannot

A region estimate can tell you which parts of the world your DNA resembles most, and in what rough mix.

It cannot tell you an exact percentage of "blood", a village, a nationality or a person's identity. A very small share is the least certain part of any estimate, so do not read much into it. A mother line and father line are separate clues, and can differ from the whole-DNA estimate.

DNA does not decide race, religion or bumiputera status. It is a statement of similarity, not of belonging.