Neuronet was taught to determine the age of people of any nationality

Scientists used data from more than 120 thousand people from Canada, South Korea and Eastern Europe to identify key indicators of aging for these populations. Then they trained the neural networks to take into account the differences in the significance of these indicators. Thus, the authors reduced the error in determining the age to a minimum: now it does not exceed six years.
With the help of machine learning, scientists have created a computer algorithm for determining the age of a set of blood test indicators. While the system can predict only a chronological age, but determining the relationship between blood counts and age is an important step on the way to an accurate estimate of the biological age.
The algorithm uses 20 biomarkers: from the concentration of glucose or hemoglobin to the number of erythrocytes in the blood. All of them are defined simply and, as a rule, are included in a standard set of analyzes in most countries.
Also, the authors identified key biomarkers, which influence aging more than others. Among them - albumin and glucose. These data are consistent with what was previously known about the change in these indicators with aging.