Researchers from Michigan State University developed a new computer tool for analyzing DNA to predict genetic traits. The tool identified gene locations connected to traits like height, bone density, and cognitive ability--and they got surprisingly accurate results.
The predictions were based on genomes from about 500,000 adults obtained from the UK Biobank, a non-profit resource supported by the UK's National Health Service.
The researchers analyzed each genome at a million separate locations called single nucleotide polymorphisms (SNP). The SNPs are not a complete genome sequence--those would cost about $1000 each. Instead, they are a map of the most common DNA variations between individuals. The data was analyzed on the MSU supercomputer, and used to train the software tool, known as an algorithm. Through the process of machine learning, the algorithm identified gene locations linked to the genetic traits to create a predictor. Each of the predictors for the three traits is based on 20,000 to 30,000 SNPs.
Lead researcher Stephen Hsu says the tool could be used to predict health risks or to create personalized drug therapies. It could also be used to screen embryos and create "designer babies." That means, eventually parents could choose the traits of their baby.
"None of this is now science fiction. Everything I'm saying to you is possible with existing technology. And so societies will have to decide, is it okay? But, all of this is going to hit us very fast," says Hsu.
As an example, Hsu said a couple screening embryos for an IVF procedure might be able to screen them for risk of a disease like type II diabetes. "Is that okay to make that decision, or is that evil? Reasonable people could have different answers," says Hsu.
The study was published in the October issue of Genetics.
