A new computer model could accurately predict a person’s height to within one inch just by analyzing their DNA

AI-driven diagnostic tools are undeniably on the precipice of revolutionizing how doctors treat and manage patients. The ability for machine-learning algorithms to crunch immense volumes of patient data and find patterns not visible to the eyes of human clinicians is revealing new ways to predict everything from breast cancer risk to a person’s chances of developing Alzheimer’s disease.

Now, a team of scientists from Michigan State University claims to have built a computer algorithm that can analyze a person’s complete genome and accurately predict how tall they are with only around a one-inch (2.5-cm) margin of error. The machine-learning system was trained on a dataset of nearly 500,000 adults.

“The algorithm looks at the genetic makeup and height of each person,” says Stephen Hsu, lead investigator on the project. “The computer learns from each person and ultimately produces a predictor that can determine how tall they are from their genome alone.”

Two other outcomes were also programmed into this early proof-of-concept project: bone density and ultimate level of education attained. While the height predictor was the most accurate, the other two outcomes still delivered reliable results. Bone density predictions were accurate enough to identify subjects most prone to developing osteoporosis associated with very low bone density.

The strength of using a computer model to crunch the data allows the system to calculate overall risk for conditions based on tens of thousands of genetic variations. Instead of current genetic testing models, which simply examine a small handful of genetic variations, this kind of big data model can identify unique genomic patterns that humans simply do not have the capacity to identify.

The researchers suggest the accuracy of the algorithms will only improve with time and larger datasets. A broad array of serious illnesses could also be included in the model as more is learnt about our complex genetic architecture.

“While we have validated this tool for these three outcomes, we can now apply this method to predict other complex traits related to health risks such as heart disease, diabetes and breast cancer,” says Hsu. “This is only the beginning.”

The increasing fusion between medical research and computer science has given rise to a new collaborative kind of research called precision health. The goal is to harness modern technologies to generate new diagnostic tools that allow early detection of disease, shifting healthcare from treating symptoms after they appear to more preventative strategies. MSU is particularly focused on this new movement, with a dedicated Precision Health Program.

Via Newatlas