What if a simple test could predict your future health risks and allow your physician to intervene now, possibly preventing or delaying the onset of illness?

Patients are one step closer to such precision health methods thanks to a recent breakthrough in predictive genomics by a research team at Michigan State University. Their findings are published in the August issue of Genetics and will be a highlight in the October issue. http://www.genetics.org/content/early/2018/08/27/genetics.118.301267

Led by researchers in computation, epidemiology and biostatistics, and physics and astronomy, the MSU group used advanced methods from machine learning to analyze almost half a million genomes.

They produced, for the first time, accurate genomic predictors for extremely complex traits such as height, bone density, and educational attainment, using United Kingdom adults.

The MSU high dimensional statistics methods can also be used to estimate how much genomic data is required to “solve” other complex traits, such as disease risks for heart disease, diabetes, breast cancer, or hypothyroidism. The authors anticipate rapid breakthroughs in the coming years, resulting in clinically useful predictors for a variety of diseases. They, and other research groups, have already produced genomic risk predictors for a number of health conditions. These predictors can identify individuals who are at many times the usual risk, using their DNA alone.

“While we have validated our computational tool for three outcomes, we can now apply these methods to predict many other complex traits,” stated Stephen Hsu, lead investigator on the project. “This is only the beginning. What we need now is more data because larger, diverse data sets will further validate the techniques, and help us to map out the genetic architecture of important traits and disease risks.”

This line of investigation differs from traditional genetic testing, which typically looks for one specific mutation that may indicate risk of future disease—Huntington’s, or BRCA in breast cancer for instance. Instead, the model considers many, many genomic indicators as well as multiple combinations to determine what risks might be present.

For their primary dataset, the MSU team used the UK Biobank cohort, comprised of almost 500k individual genotypes with multiple phenotypes. They also leveraged other datasets and single nucleotide polymorphisms (SNPs) found in earlier Genome-Wide Association Studies (GWAS) for out-of-sample validation of their results.

This type of work was thought to be five to ten years in the future, however, with the advent of machine learning and greater computing power, that time frame is beginning to shrink. Additionally, the cost to sequence a human genome is decreasing exponentially, allowing technology and medicine to converge like never before. It is not unrealistic to say now that most innovative health systems in the US will move to universal “standard of care” genotyping within a few years.

“Our team believes this is the future of medicine,” stated Hsu. “For the patient, the test is as simple as a cheek swab at a cost of about $50. Once we calculate the predictors for genetically based diseases, early intervention can save billions of dollars per patient in treatment costs, and, more importantly, save lives.”

A theoretical physicist, Dr. Stephen Hsu serves MSU as its Vice President for Research and graduate studies.

In addition to Hsu, other researchers on the project included Louis Lello, Steven G. Avery, Ana I. Vazquez, and Gustavo de los Campos (all of MSU), and Laurent Teller of the University of Copenhagen.

–Melanie Kauffman, Stephen Hsu