Neural network for diagnosing aging rate improved by UNN scientists
Lobachevsky University scientists have improved a neural network for diagnosing the rate of aging. The new immunological clock model is called SImAge (Small Immuno Age). It is based on the FT-Transformer deep neural network model. The neural network assesses the state of the human organism by 10 biomarkers that reflect the risk of age-associated diseases: heart pathologies, physical and cognitive activity disorders and others. Based on these markers, the model makes a conclusion about failures in the immune system and calculates the immunological age of a person. The new model makes it possible to test the effectiveness of therapeutic approaches, make recommendations on lifestyle adjustments, and carry out personalised diagnosis of age-related anomalies in the human organism.
"Immune cells get rid of biological debris, pathogens or unnecessary substances. This is a systemic inflammation that we don't notice. However, these processes become more active as we age, and immune cells can cause damage to healthy cells and tissues. To assess the risk of such anomalies, scientists choose different sets of biomarkers. In our new study, we have chosen 10 of the most significant ones. These are cytokines - molecules responsible for inflammatory processes and communication between immune cells. Taken together, they provide the greatest amount of information to describe a person’s immuno-inflammatory status," said Mikhail Ivanchenko, Deputy Director of the Institute for Biology of Aging, leading researcher at the UNN Laboratory of Systems Medicine of Healthy Aging.
Such models, "immunological clocks", are being developed by scientists all over the world. In their research, Nizhny Novgorod researchers are working in partnership with the University of Bologna (Italy) with the participation of the leading gerontologist Claudio Franceschi, the author of the concept of "inflammatory aging".
According to the authors of the study, SImAge is the most stable and compact "immunological clock" model to date. In the course of the study, it showed accelerated aging for patients with end-stage renal chronic disease. SImAge estimated that their immunological age exceeded their actual chronological age by 10 to 30 years.
"The classical approach, when we take tests and the indicators are compared with average values, is effective in acute cases, when symptoms are already visible and it is possible to make a diagnosis. Our method is more sensitive to inflammatory processes. It will help to assess deviations within the boundaries of the norm, to predict the appearance of age-related diseases at an early stage. This is an additional diagnostic tool, which will make the assessment of inflammatory processes more objective," explained Mikhail Ivanchenko.
The authors plan to expand the database, test the SImAge model on participants with different age-related and immunological diseases, and refine the neural network to provide a more accurate diagnosis of inflammatory processes for the widest possible range of people.
The research is part of the federal project "Artificial Intelligence" with grant support from the Analytical Centre under the Government of the Russian Federation and the Ivannikov Institute for System Programming of the Russian Academy of Sciences.
The results have been published in the international scientific journal Frontiers in Immunology.
