Method for constructing biological age prediction model incorporating social factors

Abstract

The invention discloses a method for constructing a biological age prediction model incorporating social factors. The method comprises the following steps: acquiring the age, gender, race, hematuria index, body measurement index and social factor original data of a target group; dividing into two groups of male and female original data according to genders; performing cleaning, quality control and standardization on the male and female original data to obtain male and female standard data; respectively carrying out feature screening on hematuria indexes and body measurement indexes of men and women; using a deep neural network algorithm to construct a model for standard data after male and female screening; and calculating an age acceleration value according to a model prediction result, and evaluating the aging degree. According to the method, on the basis that a biological age prediction model is constructed by using biological and genetic factors traditionally, social factors such as education level, economic condition, social support and living environment are introduced, and a more comprehensive aging evaluation mode is provided. The method comprehensively considering various influence factors is beneficial to understanding aging differences among individuals under different social backgrounds, so that health and fairness are promoted.

Description

Technical Field

[0001] The invention relates to a method for constructing a biological age prediction model incorporating social factors. Background Art

[0002] The aging of the global population is becoming increasingly serious, which not only poses a challenge to the social welfare system, but also puts forward new requirements for public health policies. Due to its complexity and the huge demand for social resources, the health problem of the elderly has become a major challenge that needs to be solved urgently. Aging is the main risk factor for chronic diseases in humans, and the essence of chronic diseases in the elderly lies in the aging and degeneration of organs. In this regard, aging biomarkers play a key role. These biomarkers are physiological and molecular indicators that indicate the structural or functional degeneration associated with aging at the level of the body, organs, tissues, cells and even subcellular levels. They can not only be used to monitor and evaluate the biological changes associated with aging, but also predict the degree to which organs evolve into pathological changes. The study of these markers helps to answer the important scientific question of "how old are we actually?" and provides the basis and premise for the assessment and early warning of aging and the intervention of aging and related diseases.

[0003] Biological age assessment is a method of evaluating an individual's physiological age and health status by measuring and analyzing the state of specific indicators or physiological processes in an organism. Compared with actual age, biological age can provide more accurate health assessment and disease risk prediction, and the difference between biological age and actual age, called "age acceleration", reflects changes in the speed of aging in the past. However, current biological age prediction models focus on biological and genetic factors, mainly using various omics data to build models, and there is no model that incorporates social factors as features. The social determinants of health (SDOH) are non-medical factors that affect health outcomes. They are the conditions under which people are born, grow, work, live, and age, and are also the broader forces and systems that shape the conditions of daily life. Adverse health behaviors, poor mental states, and social isolation can affect health and accelerate aging, and social and environmental conditions can affect the aging process.

[0004] Based on the above background, our inclusion of social factors in the construction of biological age prediction models is an attempt to comprehensively evaluate the individual aging process, which not only considers traditional biological and genetic factors, but also factors such as education level, economic status, social support and living environment. Summary of the invention

[0005] The present invention aims to improve the accuracy of the aging clock in predicting individual health risks and life expectancy. By taking into account the individual's social background, this approach promotes health equity, especially in understanding and addressing aging differences between different groups.

[0006] The technical solution adopted by the present invention is: a method for constructing a biological age prediction model incorporating social factors, comprising the following steps: (1) Obtaining original data on age, gender, race, blood and urine indicators, body measurement indicators, and social factors of the target group; (2) The raw data were divided into two groups according to gender, and the raw data of men and women were cleaned, quality controlled, and standardized to obtain standard data of men and women; (3) Characteristic screening of blood and urine indicators and body measurement indicators for men and women respectively; (4) Use the deep neural network algorithm to build a model using the standard data of male and female screening; (5) Calculate the age acceleration value based on the model prediction results and evaluate the degree of aging.

[0007] As a biological age prediction model incorporating social factors described in the present invention, the blood and urine indicators in step (1) include blood cell count, blood biochemical indicators, and urine biochemical indicators; the body measurement indicators include blood pressure, pulse, forced expiratory volume in 1 second, and body mass index; and the social factors include education level, income, Townsend deprivation index, degree of social isolation, air pollution, and noise pollution.

[0008] As a biological age prediction model incorporating social factors described in the present invention, the cleaning, quality control and standardization in step (2) are to remove the data of participants whose missing indicators are greater than 15% of the total number of indicators, and to complete the remaining data using the multiple interpolation method, and then standardize the interpolated data to standard data with a mean of 0 and a variance of 1.

[0009] As a biological age prediction model incorporating social factors described in the present invention, the feature screening in step (3) is to use the Boruta algorithm and the stepwise regression algorithm to screen out important indicators in the blood and urine indicators and body measurement indicators for predicting the age of men and women, and exclude irrelevant indicators.

[0010] As a biological age prediction model incorporating social factors described in the present invention, the model constructed using a deep neural network algorithm in step (4) is characterized in that a 10-fold cross-validation training model is adopted, multiple fully connected (Dense) layers are added, each layer uses a ReLU activation function, random uniform weight initialization, and L1 and L2 regularization. A Dropout layer is added after each Dense layer to reduce overfitting.

[0011] As a biological age prediction model incorporating social factors described in the present invention, the age acceleration value in step (5) is defined as the difference between the predicted age and the actual age. If the value is positive, it indicates that the aging speed is faster than normal. If the value is negative, it indicates that the aging speed is slower than normal.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a method for constructing a biological age prediction model incorporating social factors. The method for constructing a biological age prediction model incorporating social factors introduces social factors such as education level, economic status, social support and living environment on the basis of traditional biological and genetic factors, providing a more comprehensive aging assessment method. This method, which comprehensively considers multiple influencing factors, helps to understand the aging differences between individuals in different social backgrounds, thereby promoting health equity. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flowchart of a method for constructing a biological age prediction model incorporating social factors according to the present invention; Figure 2 A schematic diagram of participant screening and data processing provided by an embodiment of the present invention; Figure 3 A deep neural network model structure provided by an embodiment of the present invention; Figure 4 The relationship between the predicted age and the actual age of men provided by the embodiment of the present invention; Figure 5 The relationship between the predicted age and the actual age of a woman provided by an embodiment of the present invention; Figure 6 The distribution of accelerated values ​​of male age provided by the embodiment of the present invention; Figure 7 The distribution of female age acceleration values ​​provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] In order to better illustrate the purpose, technical solutions and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0015] The flowchart of the method for constructing a biological age prediction model incorporating social factors described in the present invention is as follows: Figure 1 shown.

[0016] The following steps are included: S1. Obtain the original data of age, gender, race, hematuria index, physical measurement index and social factors of the target group: download the original data of age, gender, race, hematuria index, physical measurement index and social factors of the participants from the UK Biobank database (https: / / www.ukbiobank.ac.uk / ). Hematuria index includes 26 blood cell counts: white blood cell count, red blood cell count, hemoglobin content, hematocrit percentage, mean corpuscular volume, mean corpuscular hemoglobin content, red blood cell distribution width, platelet count, hematocrit, mean platelet volume, platelet distribution width, lymphocyte count, monocyte count, neutrophil count, eosinophil count, basophil count, lymphocyte percentage, monocyte percentage, neutrophil percentage, eosinophil percentage, basophil percentage, reticulocyte count, immature reticulocyte ratio, reticulocyte percentage ratio, mean reticulocyte volume, mean spherocyte volume; 27 blood biochemical indicators: albumin, alkaline phosphatase, apolipoprotein A, apolipoprotein B, alanine aminotransferase, aspartate aminotransferase, direct bilirubin, urea, calcium, cholesterol, serum creatinine, C-reactive protein, cystatin C, γ-glutamyl transferase, glucose, glycosylated hemoglobin, high-density lipoprotein cholesterol, insulin-like growth factor 1, direct low-density lipoprotein cholesterol, phosphate, sex hormone-binding globulin, total bilirubin, testosterone, total protein, triglycerides, uric acid, vitamin D; 3 urine biochemical indicators: urine sodium, urine potassium, urine creatinine. Body measurement indicators include 5 items: diastolic blood pressure, systolic blood pressure, body mass index, pulse, and forced expiratory volume in 1 second. Social factors include 10 items: education level, income, Townsend deprivation index, frequency of visiting relatives, ability to talk, NO2 air pollution, PM2.5 air pollution, PM10 air pollution, average noise pollution level during the day, and average noise pollution level at night.

[0017] S2. The raw data are divided into two groups, male and female, according to gender. The raw data of male and female are cleaned, quality controlled, and standardized to obtain male and female standard data. The participant screening and data processing of the embodiment of the present invention are as follows: Figure 2As shown, first, 30,115 participants of non-white race were excluded from 502,373 participants aged 40 to 70 years in the UK Biobank, then 52,031 participants with missing indicators greater than 15% in step S1 were excluded, and finally 45,750 participants with no disease records and 30,927 participants with diseases before baseline were excluded. The remaining participants were classified by gender to obtain 158,852 males and 184,698 females. The multiple imputation function R package "mice" was used to impute the data of men and women respectively. The basic information of the target group is shown in Table 1. Finally, the imputed data was standardized to standard data with a mean of 0 and a variance of 1, and the formula is as follows:

[0018] in, is the standardized value, is the original data value, is the mean of the original data, is the standard deviation of the original data.

[0019] Table 1 Basic information of target groups

[0020]

[0021]

[0022]

[0023] S3. Feature screening of hematuria indexes and body measurement indexes of men and women respectively: Boruta algorithm and stepwise regression algorithm were used to screen the important indexes for age prediction of men and women in hematuria indexes and body measurement indexes, and irrelevant indexes were excluded. After screening, 30 hematuria indexes were selected for men, including white blood cell count, red blood cell count, mean corpuscular volume, red blood cell distribution width, platelet count, platelet hematocrit, lymphocyte percentage, monocyte percentage, albumin, alanine aminotransferase, apolipoprotein A, urea, serum creatinine, C-reactive protein, cystatin C, γ-glutamyl transferase, glucose, glycosylated hemoglobin, insulin-like growth factor 1, direct low-density lipoprotein, phosphate, sex hormone-binding globulin, total bilirubin, testosterone, total protein, triglycerides, vitamin D, urine creatinine, urine potassium, urine sodium; 4 body measurement indexes were selected, including diastolic blood pressure, systolic blood pressure, body mass index, and forced expiratory volume in 1 second. Women selected 25 blood and urine indicators, including white blood cell count, hemoglobin content, mean corpuscular volume, platelet count, neutrophil count, albumin, alkaline phosphatase, apolipoprotein A, aspartate aminotransferase, urea, calcium, cholesterol, C-reactive protein, cystatin C, glucose, glycosylated hemoglobin, insulin-like growth factor 1, phosphate, sex hormone-binding globulin, total bilirubin, testosterone, total protein, vitamin D, urine creatinine, urine sodium, and 4 body measurement indicators, including diastolic blood pressure, systolic blood pressure, body mass index, and forced expiratory volume in 1 second. Men and women overlapped 23 blood and urine indicators and body measurement indicators.

[0024] S4. Use the deep neural network algorithm to build models with the standard data of male and female screening: Based on the standard data of blood and urine indicators, physical measurement indicators and all social factor indicators after screening for male and female obtained in step S3, build biological age prediction models for male and female respectively. Use 10-fold cross-validation to train the model, add multiple fully connected (Dense) layers, use ReLU activation function for each layer, randomly and uniformly initialize weights, and L1 and L2 regularization (L1 = 0.01, L2 = 0.01). Add a Dropout layer after each Dense layer to reduce overfitting (Dropout rate = 0.1), use MeanSquared Error as the loss function, use Adam as the optimizer, the batch size (batch_size) is 128, and the training cycle (epoch) is 500. The network structure of this model is as follows Figure 3 shown.

[0025] S5. Calculate the age acceleration value based on the model prediction results to evaluate the degree of aging: The age acceleration value is defined as the difference between the predicted age and the actual age. If the value is positive, it means that the aging speed is faster than normal. If the value is negative, it means that the aging speed is slower than normal. The relationship between the predicted age and the actual age is as follows: Figure 4, Figure 5 The distribution of age acceleration values ​​is shown in Figure 6 , Figure 7 shown.

[0026] In summary, the method for constructing a biological age prediction model that incorporates social factors proposed in the present invention provides a more comprehensive way to assess aging. This method, which comprehensively considers multiple influencing factors, helps to understand the aging differences between individuals in different social backgrounds, thereby promoting health equity.

Claims

1. A method for constructing a biological age prediction model incorporating social factors, characterized in that: The following steps are involved: (1) Obtaining original data on age, gender, race, blood and urine indicators, body measurement indicators, and social factors of the target group; (2) The raw data were divided into two groups, male and female, according to gender. The raw data of male and female were cleaned, quality controlled, and standardized to obtain male and female standard data; (3) Characteristic screening of blood and urine indicators and body measurement indicators for men and women respectively; (4) Use the deep neural network algorithm to build a model using the standard data of male and female screening; (5) Calculate the age acceleration value based on the model prediction results and evaluate the degree of aging.

2. The method for constructing a biological age prediction model incorporating social factors according to claim 1, characterized in that: The blood and urine indicators in step (1) include blood cell count, blood biochemical indicators, and urine biochemical indicators; the body measurement indicators include blood pressure, pulse, forced expiratory volume in 1 second, and body mass index; and the social factors include education level, income, Townsend deprivation index, degree of social isolation, air pollution, and noise pollution.

3. The method for constructing a biological age prediction model incorporating social factors according to claim 1, characterized in that: The cleaning, quality control and standardization in step (2) are to remove the data of participants whose missing indicators are greater than 15% of the total number of indicators, fill in the remaining data using the multiple interpolation method, and then standardize the interpolated data to standard data with a mean of 0 and a variance of 1.

4. The method for constructing a biological age prediction model incorporating social factors according to claim 1, characterized in that: The feature screening in step (3) is to use the Boruta algorithm and the stepwise regression algorithm to screen out the important indicators of blood and urine indicators and body measurement indicators for predicting the age of men and women, and exclude irrelevant indicators.

5. The method for constructing a biological age prediction model incorporating social factors according to claim 1, characterized in that: The model constructed using a deep neural network algorithm in step (4) is characterized in that a 10-fold cross-validation training model is used, multiple fully connected (Dense) layers are added, each layer uses a ReLU activation function, random uniform weight initialization, and L1 and L2 regularization. A Dropout layer is added after each Dense layer to reduce overfitting.

6. The method for constructing a biological age prediction model incorporating social factors according to claim 1, characterized in that: The age acceleration value in step (5) is defined as the difference between the predicted age and the actual age. If the value is positive, it means that the aging speed is faster than normal. If the value is negative, it means that the aging speed is slower than normal.

Images