Aging Index
Machine learning–based system predicts life expectancy - so far only mice
Sergey Shapiro, XX2 century
The Harvard School of Medicine has developed an artificial intelligence system that can predict the lifespan of mice based on various "aging indicators" of the body. The technique can be used for rapid assessment of the impact of various surgical interventions and genetic characteristics on life expectancy and, ultimately, for the development of medical and therapeutic tools to slow down aging.
An article based on the results of long-term research at the Harvard School of Medicine was published in Nature Communications in free access (Schultz et al., Age and life expectancy clocks based on machine learning analysis of mouse frailty). At the first stage, the authors tracked the health status of 60 laboratory mice until their natural death. For more than a year, measurements were made of various indicators of the health and physical shape of mice, for example, loss of visual acuity and hearing, gait, curvature of the back, etc. This training data set was used to develop two machine learning models. One of them determines the biological age of the mouse on the basis of "aging indicators", the other is used to predict the remaining lifespan of the organism. Further observations showed that the models give a forecast with an accuracy of up to two months.
Aging indices (Frailty Index, FI) are some complex indicators of deterioration of health with age. They are non-invasive, that is, they do not affect the object and do not require excessive resources to use. For mice, there are adapted methods for determining the physical condition and calculating aging indicators. However, it is not known how accurately certain varieties of them determine the state of the organism as a whole and predict the time of its life. The FI coefficient for mice is calculated from 31 health assessment points, each of which is assigned a value of 0, 0.5 and 1 (1 denotes the corresponding age–related physical disability, 0 - its absence). Among such indicators, for example, the strength of the paw grip, the elasticity of the tail, the weakening of vision/ hearing compared to a young mouse, etc. Biologists who are interested or working in this field can study this technique more deeply from the video from the authors of the article.
Health and fitness indicators of mice used to determine the "aging index". Figures from the article by Schultz et al.
The used machine learning technique is based on decision trees – constructions in the form of tree (branching) graphs, in the nodes of which there are individual parameters that make up the FI index. Such trees are created on the basis of a training set, that is, all measured parameters for mice and real data on their life expectancy, and then allow you to make predictions of life expectancy for new individuals based on their health indicators. The authors applied an improved algorithm for learning on decision trees called random forest, or a classifier based on an ensemble of decision trees. Its essence lies in the fact that not one decision tree is used, which may not be completely accurate on a small sample, but several (a large number) such trees, each of which is built on a random sample of the original data set. In this case, the predictive model is constructed in the form of averaging over a set of such trees (a committee of decision trees), which allows improving the quality of predictions. So, the algorithm used 1000 such decision trees based on the same number of random samples of training data.
Examples of two different decision trees based on random samples of training data.
The main intended application of the developed model is to assess the impact of various clinical and therapeutic effects that presumably increase life expectancy – for example, a diet or a drug. Longitudinal studies in this case would have to last about three years (average mouse lifetime) for any drug. Therefore, such predictive biometrics is a tool that allows you to reduce the time of research, having received at least preliminary conclusions about how promising the chosen direction is.
At the next stage, predictive models were tested on groups of mice on which the effect of life-prolonging enzymes with the measurement of the FI index was previously tested. The researchers claim that the proposed machine learning model was able to correctly predict the effectiveness of appropriate therapy to increase life expectancy.
The AI system also allowed us to conclude that certain indicators of aging are more correlated with the state of health and prospects for the mouse's life expectancy in the future. For example, the degree of hearing loss and body tremor were more associated with biological age than vision loss and hair loss. It should be emphasized once again that these findings apply only to laboratory mice. So far, artificial intelligence does not know how to predict a person's life expectancy. Much more decisive factors come into play here than for mice, and they turn out to be much more interconnected.
Similar indicators of aging FI exist for humans. Moreover, the FI technique itself for mice is an adaptation of indicators originally developed for humans. But biologists do not have a reliable sample of data with such systematic monitoring of the health of people aged 60 to 90 years, including data on mortality. However, in the future they expect to develop a similar machine learning system for rapid assessment of life expectancy and the effectiveness of various therapeutic measures to prolong life.
Researchers working in this field and experimenting with mice, the authors of the work suggest using their tool on a special website- the FI calculator and the corresponding indicators of the biological age and expected life time of the mouse predicted by AI models.
Portal "Eternal youth" http://vechnayamolodost.ru