jd's Journal: Consumer Genetics, the current state of play

Ok, so let's start be defining a few terms, as it is obvious from Facebook genetic genealogy groups that people are truly ignorant on the subject. (Not that I believe this is common on Slashdot, where we're all much more knowledgeable.)

First off, most genetic testing is NOT carried out by sequencing all of your DNA, a widespread belief that resulted in outrage on one Facebook group when I pointed that out.

The vast majority of consumer testing is done by SNP genotyping. They look at very specific genetic markers and see if those markers have changed from one base pair to another. That's the only type of mutation looked for and they typically look at only a few.

So we've our first way to group companies: sequencing vs genotyping.

SNPs (single nucleotide polymorphisms) are, as mentioned above, one type of mutation. Another is called STR (short tandem repeat), where a block of DNA is duplicated.

FamilyTreeDNA does both STR and SNP testing, STRs mostly for the Y chromosome. Both can be used for family history.

Most labs, though, use only SNP tests. It's quicker and cheaper than counting repeats but with many of the more interesting ones covered by patents or kept private by other means, there's a lot more secrecy involved.

(Note: This has doubtless led to a lot of unnecessary deaths, as genetic markers indicating a high probability of getting certain forms of cancer are being milked by private companies for profit. Few people get more than one test, so most people won't know if they carry such markers and can't take action in advance.)

So the second piece of jargon is SNP vs STR.

Finally, we come to the different areas of DNA. There are regions that are especially good for ancestrial reserch (mostly non-coding DNA), then there's the exome (which is where most of the protein coding takes place), you've telomeres (suicidal buffers between chromosomes, which have a function in longevity), and so on. I won't list them all.

The Y chromosome is particularly good for ancestry, but only has 9 coding genes left in it. It's possible it will vanish in time, but it seems to be fairly stable for right now.

Most companies test only DNA that is good for ancestral research in the autosomal regions (aDNA, the regions outside the sex chromosomes). This allows you to identify anyone who is genetically connected, but because you (on average) get just under 50% (remember, there's mutations in each generation and that DNA comes from neither parent) of your DNA from each parent, the distance you can track depends on how many markers are tested (very few). Reliability falls off sharply.

YDNA (Y chromosome DNA) tests only test for paternal ancestry, but if two people have a common paternal-line ancestor, it's a lot more precise once you're past about second cousins. It's popular with anthropologists as it's very good for tracking how men have migrated.

mtDNA (mitochondrial DNA) is only inherited through the maternal line. Again, it's very popular, this time for tracking how women have migrated. There are certain forms of mtDNA that are linked to health benefits and others to genetic diseases, so this one tends to be the most controversial of the ancestral DNA tests. It also changes very slowly, so you don't get high resolution on population movements.

These two (YDNA and mtDNA) tests can tell you a lot about whether societies are open or closed, and whether it was men who travelled to find partners, women, or both. So we can know something of the culture of even long-extinct societies.

The data I have been able to find is for 2019. It shows: Myheritage tests for 702,442 autosomal SNPS, AncestryDNA for 637,639, FTDNA for 612,272 and 23andme for 630,132. This is out of a total of 3 billion base pairs. So the best test that year looked at 0.0023% of the genome.

ISOGG produced a chart as well, but it's far older. Their chart is dated around 2013.

Since you inherit a random 50% from each parent, the assumption that this is statistically meaningful for such a small fraction of the DNA is questionable. It seems to work adequately, but I'm not sure what the error bars are.

FTDNA also tests up to 111 STRs on regular tests and 600+ STRs for their "BigY" (it depends on the quality of the genetic sample).

Companies that do sequencing sometimes offer partial kits (in the order of tens of millions of SNPs) or full sequencing (which is what the same suggests). These are rarer and more expensive.

Most DNA companies allow you to access the raw data, some only allow it if you pay vast sums of money, and some don't allow you to at all. Always check in advance.

When you download your own data, you can use public databases to search for matches (either for relatives or genetic conditions). The quality of public databases is less controlled, both in terms of privacy and quality of data. However, corporate databases will usually be smaller for both types of data and will also usually not contain data from rivals. If you want broad data sets, public databases are the way to go.

I've only tested with 23&Me, FamilyTreeDNA, CRI Genetics and Nebula Genomics, so can't tell you anything much about the quality of the other companies.

(Ok, I also tested with uBiome, a microbiome testing company in the US, but they had their computers seized some time back due to fraud. I have no idea what happened to my data on there, or whether there's a way to access it.)

The quality seems to be reasonable for all four.

FTDNA is the most expensive for a lot of things, but has less of a sticker shock than Nebula and gets more data than 23&Me. It looks like there are a few companies that are better for ancestry but it's one of the best and the one the Genomics Project used. They're the only ancestral company that gives you STRs AFAIK and they give you a much more detailed evaluation of haplogroups than anyone else I've tested with.

Nebula does up to medical grade (100x oversampling) DNA testing, so if you want results a hospital will trust, that's where you part with a vast amount of money.

23&Me is good for a lot of medical stuff and if you want to help with research is probably the best.

CRI Genetics produces a lot of data with much higher reliability than most of the others, but you can't access the raw data and their databases won't be as extensive. However, because you can't access the raw data, you have to test with them to compare against their database.