I have posted a few notes about predictive diagnostic algorithms (see: An Algorithm Using Medical Record Data Predicts Risk for Parkinson's Disease; Should the Work Product of "Non-Explainable" Medical Algorithms Be Ignored) and disease risk scores (see: Genetic Risk Scores: Ready for Inclusion in the Medical Record?). Using AI techniques, researchers at Massachusetts General are developing a web site that will display a risk score for five common diseases after consumers upload their DNA to the site (see: Clues to Your Health Are Hidden at 6.6 Million Spots in Your DNA). The algorithm takes into account millions of alterations in the DNA rather than a few focal changes. Below is an excerpt from the article about this news:
...[R]esearchers are now building a website that will allow anyone to upload genetic data from a company like 23andMe or Ancestry.com. Users will receive risk scores for heart disease, breast cancer, Type 2 diabetes, chronic inflammatory bowel disease and atrial fibrillation....Dr. Sekar Kathiresan, senior author of the new paper and director of the Center for Genomic Medicine at Massachusetts General Hospital, said his team had validated the heart risk calculation in multiple populations. But DNA is not destiny, Dr. Kathiresan stressed. A healthy lifestyle and cholesterol-lowering medications can substantially reduce risk of heart attack, even in those who have inherited a genetic predisposition. The new tool also can find people at the low end of the risk range for the five diseases. This should prove useful to certain patients: for example, a woman who is trying to decide when she should start having regular mammograms, or a 40-year-old man with a slightly high cholesterol level who wants to know if he should take a statin....
“It carries great hope, but also comes with a lot of questions,”....“Who should get tested? How should the results be provided? ...[W]ill the results actually lead people to make decisions that improve their health? People may need genetic counseling before and after getting these sorts of risk scores....Patients may not appreciate the consequences of learning they have a high likelihood of having a heart attack or breast cancer or one of the other diseases the test assesses.....But medical experts said this sort of risk assessment is the wave of the future....The study began because there was general agreement among researchers that many common diseases are linked not to one mutation, but rather to thousands or millions of mutations....In recent years, scientists have cataloged more than 6 million tiny changes in DNA that slightly affect the chances that people will get various diseases.....But it should be possible, scientists felt, to combine data on all the small DNA changes to construct an individual risk score. To do that, the researchers needed a new algorithm that would weigh the significance of the variations in the genes....The investigators found that their algorithm did predict the odds of being diagnosed with one of the five diseases.
I personally believe that predictive analytic web sites such as the one described above will provide significant public health benefits. But, as always, there will be downside risks. The most obvious is that some individuals informed about having a risk for a particular disease such as IBD may obsess about this possibility. On the more positive side, the five diseases listed above (heart disease, breast cancer, Type 2 diabetes, chronic inflammatory bowel disease and atrial fibrillation) constitute a large percentage of the chronic disease burden in the U.S. If the users of the above mentioned web site take serious measures to forestall the onset of a particular disease (e.g., weight loss for type 2 diabetes), the consequences can be very significant.
I have already personally gained substantial benefit from having my DNA analyzed by 23andMe in terms of my learning about an inherited disease for which I am a carrier and also support for my hobby of genealology. Hoever, the vast majority of our population will not have the means or an interest in obtaining disease risk scores using AI technology. However, the future for predictive diagnostic analytics is now becoming quite clear. We will initially see a number of web sites such as the one described above that will perform such services for a fee. In the long run, such predictive analytics will be performed in our health system labs and then offered as a routine health test. Appropriate and necessary monitoring for the appearance of particular disease(s) will then be performed in that health system.