AI can do a lot of gitics
Artificial intelligence predicts human growth and health risks based on DNA analysis
Yulia Vorobyova, Vesti
Modern diagnostic tools, which are based on the work of artificial intelligence, promise a real revolution for personalized medicine. The indisputable advantage of such systems is the ability to process huge arrays of data, as well as fast learning and high accuracy of analysis.
New algorithms can already cope with diagnostics at the medical level and even give the best result. But, more importantly, they can be used for tasks that clinicians cannot do.
One example is a new development by American researchers. This is an algorithm capable of analyzing the complete human genome and predicting with high accuracy its growth, bone density, level of education, and – in the future – some health indicators.
As noted in the press release The new DNA tool predicts height, shows promise for serious illness assessment by the authors of the work from the University of Michigan, this is the first tool of its kind.
Experts explain: as a rule, traditional tools are aimed at finding specific changes in genes and chromosomes, which, for example, may indicate an increased risk of developing a particular disease.
Meanwhile, the new system analyzes numerous genomic data and builds a predictor (predictive parameter) based on tens of thousands of genomic variations. A person cannot do such a job.
At the first stage, the researchers trained the system to work with three predictors – human height, bone density and education level.
The developers applied machine learning methods and used the data of about half a million people stored in the UK Biobank database to train the algorithm. All the system needed was the DNA of each participant.
Validation tests followed, and the computer "predicted" the height of each person with an error of no more than three centimeters.
The accuracy of forecasting other parameters was sufficient to use these data to identify patients with an increased risk of developing various diseases (for example, osteoporosis – insufficient bone density), as well as people who are highly likely to encounter (have encountered) with difficulties at school.
At the next stage, scientists plan to improve the algorithm and increase its accuracy during training on larger samples.
Another area of work is to train the system to detect a wider range of diseases based on genomic data. Among the first targets are diabetes mellitus, cardiovascular diseases and breast cancer.
As the lead author of the paper, Stephen Hsu, noted, with the increase in computing power of computers and the reduction in the cost of genome sequencing, research that once took five to ten years is now becoming more affordable and less expensive.
"Our team believes that this is the future of medicine. Once we calculate predictors for genetically determined diseases, early intervention will help save billions of dollars in treatment costs and, more importantly, save lives," the researcher concludes.
A more detailed description of his team's work is contained in an article published in the journal Genetics (Lello et al., Accurate Genomic Prediction of Human Height).
Let's add that forecasting growth by DNA is by no means an "advertising move" of specialists. Growth for doctors is one of the important parameters with which health risks may be associated. Men with high growth are more likely to develop prostate cancer. Other studies have shown that the relationship in a couple and even life expectancy largely depend on growth.
We also recall that earlier neural networks taught doctors to diagnose skin cancer as well as doctors, predict the behavior of people with mental disorders and determine the risk of death of a patient.
Portal "Eternal youth" http://vechnayamolodost.ru