23andme uses GWAS SNPs, about 1 million of them. You have a 3 billion base pair haploid genotype, so they sample ~1/3000 base pairs. What does this mean? First

23andme uses GWAS SNPs, about 1 million of them. You have a 3 billion base pair haploid genotype, so they sample ~1/3000 base pairs. What does this mean? First, most of the SNPs are noncoding, found not in exons but in introns and intergenic space. Still, you can't dismiss these; some are causal for major gene expression changes (my lab has articles set to be published next week on this topic; more details then). Second, the fact that you're only covering 1/3000 nucleotides tells you that there's no way you can fully specify someone's genetic risk with this data, ever. There is a debate in the community right now about whether common variants (which are on 23andme type chips) or rare variants contribute most of the genetic variability. I think it's clear that common variants are winning this game — but this is on a population level! On an individual level, I don't think I'd ever be comfortable telling someone their risk of X with just GWAS data. For the population, their rare/private mutation in Gene Y has little impact, perhaps, but for try telling that to the 5 people with the Gene Y mutation who will die by age 20. (Extreme hypothetical to try to drive home the broader point.) Let's also not forget that phenotype is a computed expression of genotype+environment. If your genetic risk score from 23andme says you are at high risk of heart disease, yet your grandparents on both sides are 100 and healthy and your LDL-C is 60, should you really be concerned?