Biological age evaluation system based on clinical indexes, construction method and application

Abstract

The invention relates to a clinical index-based biological age evaluation system, a construction method and application, and belongs to the field of biological aging research. Firstly, clinical indexes of a subject are collected, a biological age estimation model is constructed based on the clinical indexes, and the biological aging speed is further calculated. The method can be used for identifying an individual which accelerates senescence or delays senescence, and risk prediction and crowd risk stratification of death and senescence related diseases are realized. According to the method, the nonlinear relationship between clinical indexes and the aging degree is considered, so that the biological age estimated by the method has a better prediction effect on the risk of death and new aging related diseases. The biological age and the biological aging speed are quantified based on the clinical indexes, measurement of the clinical indexes can be completed only through conventional physical examination and one-time blood sampling, the method has the advantages of being low in cost and convenient to use, the biological aging speed can be evaluated in time after physical examination, and health guidance opinions are provided for aging acceleration individuals.

Description

Technical Field

[0001] The present invention relates to the field of biological aging research, and more specifically, to a biological age assessment system based on clinical indicators, a construction method and an application thereof. Background Art

[0002] In order to deal with aging-related diseases and delay or reverse the aging process, it is necessary to measure biological age and identify those individuals who age faster or slower. In existing studies, easily accessible and inexpensive clinical indicators are ideal markers for constructing biological age predictors. Chen et al. constructed KDM-BA using 12,377 Chinese people from the Kadoorie Biobank in China, which estimates biological age by giving higher weights to biomarkers that are more related to age (Chen, Zhang et al. 2023). However, the biological age estimator constructed based on the supervised model with actual age as the training label captures a large amount of actual age-related variation, while providing limited information independent of actual age in predicting diseases (Galkin, Mamoshina et al. 2020). Some researchers believe that the survival time of individuals of the same actual age reflects the relative size of their biological age, and then focus on developing all-cause mortality risk predictors (Lu, Quach et al. 2019, Pyrkov, Avchaciov et al. 2021). However, existing biological age assessment methods are mainly based on European populations and do not consider the nonlinear association between clinical indicators and mortality risk. Therefore, there is a lack of biological age assessment methods in this field that are based on large prospective cohorts of the Chinese population and take into account the nonlinear association between clinical indicators and mortality risk.

[0003] In the Chinese invention patent with publication number CN 118448041A, clinical indicators are used to construct a method for estimating the rate of physiological aging. This method assumes that there is a linear correlation between each clinical indicator and the actual age and the risk of death. Based on this assumption, the patent first uses linear regression to correct the actual age for each clinical indicator, and then predicts the risk of death within a 6-year follow-up period based on the residual value of the clinical indicator after correcting the actual age. Finally, the physiological aging rate of the sample is estimated based on the prediction model. However, existing studies have shown (Nguyen, Colacino et al. 2021) that most clinical indicators are U-shapedly correlated with the risk of death, that is, values ​​above or below the clinical recommended values ​​will lead to a higher risk of death. Summary of the invention

[0004] In order to solve the problems that the existing biological age estimators lack research based on Chinese population cohorts and do not consider the nonlinear association between clinical indicators and mortality risk, the present invention proposes a biological age assessment system based on clinical indicators, a construction method and an application, with the aim of providing a biological age assessment method based on a large prospective Chinese population cohort and using easily available clinical indicators such as blood routine and blood biochemistry.

[0005] According to a first aspect of the present invention, there is provided a method for constructing a biological age assessment system, comprising the following steps:

[0006] (1) Collect clinical indicators of subjects;

[0007] (2) performing nonlinear Cox proportional hazards regression analysis on the clinical indicators described in step (1) and the risk of death;

[0008] (3) For clinical indicators that are not nonlinearly associated with the risk of death identified in step (2), the original values ​​are not converted; for clinical indicators that are U-shapedly associated with the risk of death identified in step (2), their original measured values ​​are converted based on their optimal values; the optimal value is defined as the value of the clinical variable in the restricted cubic spline (RCS) curve when the relative risk of death is the lowest;

[0009] (4) Predicting the risk of death based on actual age, the original clinical indicators obtained in step (3), and the converted variables, using the LASSO algorithm for variable screening in the Cox proportional hazard regression model;

[0010] (5) Calculate the biological age PAI of each sample based on the variables screened in step (4);

[0011] (6) Based on the biological age PAI calculated in step (5) and the actual age of the sample, calculate the biological aging rate ΔPAI of each sample.

[0012] Preferably, in step (1), the indicators are height, weight, blood pressure, blood lipids, blood sugar, routine blood test indicators and blood biochemical indicators, and the skewed distribution indicators among the indicators are normally transformed.

[0013] Preferably, in step (2), the nonlinear Cox proportional hazards regression model adopts a restricted cubic spline method, with actual age and a single clinical indicator as covariates, and the knots of the clinical indicator are selected from 10%, 50% and 90% quantiles.

[0014] Preferably, in step (3), the formula for variable conversion of the clinical indicator is:

[0015]

[0016]

[0017] Among them, y i and represent the original value and clinical optimal value of the i-th clinical indicator respectively.

[0018] Preferably, in step (4), the Cox proportional hazards regression model is expressed as:

[0019]

[0020] in, and h0(t) represent the hazard rate function and baseline hazard function at time point t, respectively, β i represents the regression effect value of the i-th indicator, y i· represents the i-th variable of all samples, and p is the number of the indicators. The covariates included in the Cox proportional hazard regression model include the actual age CA, the original values ​​of the clinical indicators linearly associated with death, and the transformed values ​​of the clinical indicators nonlinearly associated with death.

[0021] Preferably, in step (5), the PAI calculation formula of sample j is:

[0022]

[0023] Where k is the number of variables selected for calculating PAI, y i· represents the value of the i-th variable of all samples, represents the weight of the i-th variable.

[0024] Preferably, in step (5), the variable weight From multiple linear regression analysis. In this multiple linear regression analysis, if a clinical indicator has a U-shaped correlation with the risk of death, there is only one variable transformation value (y i or i2 ) is selected into the model by the LASSO algorithm, and y is retained at the same time i and i2 Used to calculate PAI.

[0025] Preferably, in step (6), the ΔPAI calculation formula of sample j is:

[0026]

[0027] Among them, x j represents the actual age of the jth sample, and represent the calculated slope and intercept terms in the linear regression of PAI on chronological age, respectively.

[0028] According to another aspect of the present invention, there is provided a biological age assessment system, comprising:

[0029] Clinical index collection module: used to collect clinical indexes of subjects;

[0030] Nonlinear effect identification module: Based on nonlinear Cox proportional hazard regression analysis, clinical indicators with a U-shaped correlation with the risk of death are identified; the value of the clinical variable when the relative risk of death in the restricted cubic spline curve is the lowest is defined as the clinical optimal value;

[0031] Nonlinear effect conversion module: For clinical indicators that have a U-shaped correlation with the risk of death, their original measured values ​​are converted based on their optimal values;

[0032] Variable screening module: Based on the candidate variables, a Cox proportional hazard regression model is constructed, and the LASSO algorithm is used for variable screening;

[0033] Biological age estimation module: Calculates biological age PAI based on the predictors obtained by the nonlinear effect identification module and the variable screening module;

[0034] Biological aging rate estimation module: Calculate the biological aging rate ΔPAI based on the biological age PAI and actual age.

[0035] According to another aspect of the present invention, a biological aging assessment method is provided, which collects the following indicators of the sample: actual age CA, in years; body mass index BMI, in kg / m 2 , defined as weight divided by the square of height, weight in kg, height in m; systolic blood pressure SBP, in mmHg; fasting blood glucose FBG, in mmol / L; total cholesterol TC, in mmol / L; red blood cell count RBC, in 10 12 / L; mean corpuscular volume MCV, in fL; mean corpuscular protein MCH, in pg; coefficient of variation of red blood cell volume distribution width RDW, in %; white blood cell count WBC, in 10 9 / L; lymphocyte percentage LYM, in %; monocyte percentage MONO, in %; alkaline phosphatase ALP, in μ / L; aspartate aminotransferase AST, in μ / L, alanine aminotransferase ALT, in μ / L; direct bilirubin DBIL, in μmol / L; serum creatinine SCR, in μmol / L; urea nitrogen BUN, in μmol / L; PAI and ΔPAI of the sample are calculated according to the biological aging assessment system, so as to assess the degree and speed of biological aging of the sample;

[0036] When ΔPAI>0, it means that the biological aging rate exceeds that of peers; when ΔPAI=0, it means that the biological aging rate is equal to that of peers; when ΔPAI<0, it means that the biological aging rate is lower than that of peers.

[0037] According to another aspect of the present invention, there is provided application of the method in assessing the degree of aging and predicting the risk of death or disease.

[0038] In general, the above technical solution conceived by the present invention has the following technical advantages compared with the prior art:

[0039] (1) A method for assessing biological age applicable to the Chinese population is provided. This method is used to assess the biological age and biological aging rate of Chinese population samples. Compared with existing biological age estimators based on European populations or linear estimators using actual age as the prediction label, this method has higher prediction accuracy for the risk of death and aging-related diseases and is more capable of identifying individuals with accelerated or delayed aging.

[0040] (2) Compared with the existing estimation methods for the Chinese population, this method is based on general population data with a larger sample size and longer follow-up time during the model training phase, making the biological age estimation method obtained by this method more robust;

[0041] (3) The present invention takes into account the nonlinear association between clinical indicators and the risk of death, and provides a method for estimating biological age that simultaneously considers different risk effects below and above the clinical optimal value, so that the method can obtain a better prediction effect of biological age on the risk of death and new aging-related diseases. The present invention first identifies clinical indicators that are U-shapedly associated with the risk of death, and then uses the new method proposed in the present invention to perform variable conversion on the original measured values ​​of the clinical indicators. This conversion method enables the present invention to model the different risk effects of clinical indicators on death when they are above and below the clinical recommended values, which is more in line with the epidemiological characteristics of biological aging. On the other hand, the present invention simultaneously provides a method for estimating biological age and the rate of biological aging, which can provide more comprehensive scientific advice and guidance for aging prevention and intervention.

[0042] (4) The present invention estimates biological age and biological aging rate simultaneously based on clinical indicators. The measurement of clinical indicators only requires a routine physical examination and a blood draw, which has the advantages of being cheap and convenient. The biological age and biological aging rate can be calculated in time after the physical examination, providing health guidance for individuals with accelerated aging. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Schematic diagram of a biological age assessment method provided by an embodiment of the present invention.

[0044] Figure 2 Schematic diagram of a biological age assessment system provided by an embodiment of the present invention.

[0045] Figure 3 It is an analysis flow chart for constructing and verifying a biological age index based on a population cohort provided in an embodiment of the present invention.

[0046] Figure 4 It is a restrictive cubic spline curve between the clinical index and the risk of death provided by the embodiment of the present invention.

[0047] Figure 5 It is the joint distribution of actual age and PAI in the training set and the test set provided by the embodiment of the present invention.

[0048] Figure 6 This is the predictive effect of PAI and ΔPAI on the risk of death in the "Dongfeng-Tongji" cohort provided by an embodiment of the present invention.

[0049] Figure 7 It is the relative risk ratio of ΔPAI for predicting cardiovascular disease in the "Dongfeng-Tongji" cohort provided by an embodiment of the present invention.

[0050] Figure 8 This is the predictive effect of PAI and ΔPAI on the risk of death in the UK Biobank samples provided by the embodiments of the present invention.

[0051] Fig. 9 This is the predictive effect of ΔPAI on predicting aging-related diseases in UK Biobank samples provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0053] To achieve the above objectives, according to one aspect of the present invention, a method for constructing a biological age assessment system based on clinical indicators is provided, including a data preprocessing stage and a model determination stage, the data preprocessing stage includes S1 and S2, and the model determination stage includes S3, S4, S5 and S6.

[0054] Step S1: Collect the age and gender information of healthy subjects, and complete the measurement of height, weight, blood pressure, as well as blood lipid, blood sugar, blood routine and blood biochemistry tests. Follow up the subjects for up to 10 years and collect the information on the subjects' deaths and new cardiovascular disease events.

[0055] Preferably, the clinical indicators (abbreviations, units) include the following: body mass index (BMI, kg / m 2 ), waist circumference (Waist, cm), systolic blood pressure (SBP, mmHg), diastolic blood pressure (DBP, mmHg), pulse pressure (PP, mmHg), fasting blood glucose (FBG, mmol / L), high-density lipoprotein cholesterol (HDL-C, mmol / L), low-density lipoprotein cholesterol (LDL-C, mmol / L), total cholesterol (TC, mmol / L), triglycerides (TG, mmol / L), red blood cell count (RBC, 10 12 / L), mean corpuscular volume (MCV, fL), red blood cell volume distribution width (RDW, %), mean corpuscular protein concentration (MCHC, g / L), mean corpuscular protein content (MCH, pg), hemoglobin (HGB, g / L), hematocrit (HCT, %), white blood cell count (WBC, 10 9 / L), monocyte percentage (MONO, %), lymphocyte percentage (LYM, %), neutrophil percentage (NEUT, %), basophil percentage (BASO, %), eosinophil percentage (EOS, %), platelet count (PLT, 10 9 / L), mean platelet volume (MPV, fL), platelet volume distribution width (PDW, fL), platelet packed volume percentage (PCT, %), aspartate aminotransferase (AST, μ / L), alkaline phosphatase (ALP, μ / L), alanine aminotransferase (ALT, μ / L), total bilirubin (TBIL, μmol / L), direct bilirubin (DBIL, μmol / L), indirect bilirubin (IDBIL, μmol / L), urea nitrogen (BUN, mmol / L), creatinine (SCR, μmol / L), uric acid (SUA, μmol / L).

[0056] Step S2: Quality control and preprocessing of the obtained sample baseline characteristics and clinical index raw data. According to the technical solution of the present invention, the data preprocessing process is:

[0057] A10, 13,555 participants from the Dongfeng-Tongji cohort enrolled in the Central Hospital in 2008 were used as the training set; 26,990 participants from the Dongfeng-Tongji cohort enrolled in the Central Hospital, Huaguo Hospital, Maojian Hospital, and Xiyuan Hospital in 2013 were used as the test set. Samples and clinical indicators were filtered for the samples in the training set and validation set, respectively. Finally, 12,769 participants were included in the training set, 15,904 participants were included in the validation set, and 36 clinical indicators were included.

[0058] A20, samples with baseline cancer or missing variables >5 were excluded.

[0059] A30, because the measured values ​​of some clinical indicators showed obvious skewed distribution, natural logarithm transformation was performed on triglycerides, fasting blood glucose, monocyte ratio, eosinophil percentage, basophil percentage, red blood cell volume distribution width, mean hemoglobin concentration, mean platelet volume, aspartate aminotransferase, alkaline phosphatase, alanine aminotransferase, total bilirubin, direct bilirubin, indirect bilirubin, urea nitrogen and creatinine.

[0060] Preferably, samples with extreme outliers in any variable are excluded, where extreme outliers are defined as 8 interquartile ranges (IQR) away from the median.

[0061] A40, missing variables of each sample were filled based on the multiple imputation (PMM) method.

[0062] Step S3: In the training set, the restricted cubic spline method is used to identify the nonlinear association between clinical indicators and mortality risk. In the Cox proportional hazard model, the actual age is corrected and the 10%, 50%, and 90% quantiles are set as nodes of clinical indicators.

[0063] Step S4: Determine the optimal value of the clinical indicator based on the U-shaped association identified in step S3 is the lowest point of the RCS curve. Based on this optimal value The original values ​​of clinical indicators are converted into two new variables:

[0064]

[0065]

[0066] Among them, y i and Represent the original value and clinical optimal value of the i-th clinical indicator respectively. For clinical indicators with a U-shaped correlation with the risk of death, this transformation method can simultaneously convert values ​​above and below For clinical indicators that are linearly associated with the risk of death, no conversion is performed on the original values ​​of the clinical indicators.

[0067] Step S5: Cox proportional hazard regression model and LASSO method were used to select clinical indicators with predictive value for mortality risk. Based on the multivariate linear regression model and the screened predictive factors, a physiological aging index (PAI) was constructed.

[0068] Preferably, LASSO is used for variable screening in the Cox proportional hazard regression model, and the Cox proportional hazard regression model is expressed as:

[0069]

[0070] in, and h0(t) represent the hazard rate function and baseline hazard function at time point t, respectively, β i represents the regression effect value of the i-th indicator, y i· represents the i-th variable of all samples, and p is the number of the indicators. The covariates included in the Cox proportional hazard regression model include the actual age CA, the original values ​​of the clinical indicators linearly associated with death, and the transformed values ​​of the clinical indicators nonlinearly associated with death.

[0071] Preferably, k variables with predictive value for mortality risk are selected based on LASSO, and PAI is defined as the weighted sum of k variables:

[0072]

[0073] Where k is the number of variables selected for calculating PAI, y i· represents the value of the i-th variable of all samples, represents the estimated value of the regression effect of the ith variable.

[0074] Preferably, the variable weights From multiple linear regression analysis. In this multiple linear regression analysis, if a clinical indicator has a U-shaped correlation with the risk of death, there is only one variable transformation value (y i or i2 ) is selected into the model by the LASSO algorithm, and y is retained at the same time i and i2 Used to calculate PAI.

[0075] Step S6: Based on the PAI calculation formula determined by the training set samples, calculate the PAI and ΔPAI of each sample in the test set. ΔPAI is defined as the residual value of the regression of PAI on actual age CA. The calculation formula of ΔPAI of the jth sample is expressed as:

[0076]

[0077] Among them, x j represents the actual age of the jth sample, and They represent the slope and intercept terms calculated in the linear regression of PAI on actual age. The ΔPAI of the test set samples was standardized to have a variance of 1.

[0078] The predictive effect of PAI on mortality risk was verified based on the Cox regression model:

[0079] h(t,d j )=h0(t)exp(γPAIj )

[0080] Among them, h(t,d i ) and h0(t) represent the hazard rate function and baseline hazard function of sample j at time point t, respectively. i represents the survival status of sample j at time point t, PAI j represents the biological age of sample j at the baseline time point. γ means that for every 1 standard deviation increase in PAI, the relative mortality of the sample increases by “exp(γ)–1” times.

[0081] According to the model determination method provided in steps S3, S4, S5 and S6, the biological age (PAI) and biological aging rate (ΔPAI) of an individual can be quantified. ΔPAI>0 indicates that the individual is an accelerated aging individual, and ΔPAI<0 indicates that the individual is a delayed aging individual. This indicator can provide information independent of actual age for predicting the risk of death and new aging-related diseases, and has broad application prospects and market potential.

[0082] According to another aspect of the present invention, a biological age assessment system based on the above method is provided, the system comprising the following modules:

[0083] Data collection module: Collect basic information of healthy participants, including gender, age, and baseline medical history. Participants undergo physical examination and laboratory tests in the hospital after fasting for eight hours to collect clinical indicator information required for biological age assessment.

[0084] Nonlinear effect identification module: Based on nonlinear Cox proportional hazard regression analysis, clinical indicators with a U-shaped correlation with the risk of death are identified; the value of the clinical variable when the relative risk of death in the restricted cubic spline curve is the lowest is defined as the clinical optimal value;

[0085] Nonlinear effect conversion module: For clinical indicators that have a U-shaped correlation with the risk of death, their original measured values ​​are converted based on their optimal values. The variable conversion calculation formula is:

[0086]

[0087]

[0088] Among them, y i and Represent the original value and clinical optimal value of the i-th clinical indicator respectively. For clinical indicators with a U-shaped correlation with the risk of death, this transformation method can simultaneously convert values ​​above and below For clinical indicators that are linearly associated with the risk of death, no conversion is performed on the original values ​​of the clinical indicators.

[0089] Preferably, there is a U-shaped correlation between the 25 clinical indicators and the risk of death, including: aspartate aminotransferase AST, serum urea nitrogen BUN, mean corpuscular volume MCV, mean corpuscular protein MCH, alanine aminotransferase ALT, red blood cell count RBC, serum creatinine SCR, hematocrit HCT, platelet count PLT, high-density lipoprotein cholesterol HDL-C, hemoglobin HGB, direct bilirubin DBIL, fasting blood glucose FGB, platelet volume percentage PCT, total cholesterol TC, low-density lipoprotein cholesterol LDL-C, lymphocyte percentage LYM, neutrophil percentage NEUT, alkaline phosphatase ALP, serum uric acid SUA, total bilirubin TBIL, white blood cell count WBC, body mass index BMI, monocyte percentage MONO, waist circumference Waist.

[0090] Variable screening module: Based on the candidate variables, a Cox proportional hazard regression model is constructed, and the LASSO algorithm is used for variable screening;

[0091] Preferably, according to the biological age assessment system provided by the present invention, 33 variables related to the risk of death are screened from 61 variables (11 original values ​​of clinical indicators and 50 converted values) of actual age CA and 36 clinical indicators using Cox proportional hazard regression based on LASSO, and the effect size of each predictor comes from multivariate Cox regression.

[0092] Biological age calculation module: Based on the prediction factors and their weights obtained by the nonlinear effect identification module and the variable screening module, the biological age PAI and the biological aging rate ΔPAI are calculated.

[0093] Preferably, the calculation formula of PAI is:

[0094]

[0095] Each predictor variable y i· and its weight As shown in Table 1.

[0096] Table 1

[0097]

[0098] Biological aging rate estimation module: Calculate the biological aging rate ΔPAI based on the biological age PAI and actual age.

[0099] Preferably, the calculation formula of ΔPAI is:

[0100] ΔPAI j =PAI j -0.12×x j+7.4.

[0101] Among them, x j Represents the actual age of the jth sample.

[0102] Example 1

[0103] Figure 1 For a schematic diagram of the biological age assessment method provided by the present invention, see Figure 1 , this embodiment explains the biological age assessment method based on data from the "Dongfeng-Tongji" cohort. Figure 2 For a schematic diagram of the biological age assessment system provided by the present invention, see Figure 2 This example further explains the biological age assessment system based on individual data from the UK Biobank. Figure 3 .

[0104] Operation S1, collect data and divide samples from the "Dongfeng-Tongji" cohort.

[0105] In the large prospective cohort "Dongfeng-Tongji" cohort established in the applicant's unit, questionnaire surveys, physical examinations and laboratory tests were completed for all research subjects. The questionnaire survey was conducted by uniformly trained and qualified investigators using a semi-structured questionnaire to collect basic information of the research subjects, including gender, age, date of birth, individual health behavior, medical history, medication history, etc.; the physical examination was conducted by clinicians to test the height, weight, blood pressure, electrocardiogram, waist circumference and hip circumference of the research subjects; laboratory tests required the subjects to fast for more than 8 hours, and then draw blood from the elbow vein to test the blood routine, blood biochemistry, blood lipids and blood sugar of the blood samples. The follow-up working group conducted a final follow-up of the participants through the electronic medical records of the Dongfeng Motor Health Insurance Network and classified the death events according to the International Classification of Diseases, Tenth Edition (ICD-10). The follow-up deadline for death events was defined as May 20, 2019. This study does not include accidental death samples as outcomes (ICD-10 code: V01-Y98). According to the enrollment time of the participants, the samples were divided into a training set and a validation set, of which the training set contained 12,769 samples and the validation set contained 15,904 samples.

[0106] Operation S2, perform sample and variable quality control on the training set and validation set. This step corresponds to Figure 1Step 1 of the biological age assessment method proposed in the present invention. In the training set samples: ① exclude samples with cancer when enrolled in 2008; ② exclude samples with variable missing numbers greater than 5 and samples lost to follow-up; ③ perform natural logarithm transformation on variables with right-skewed distribution; ④ exclude outlier samples according to the standard of deviation from the mean 8 times the interquartile range; ⑤ use multiple imputation (PMM) algorithm to fill in the missing values ​​of samples; in the test set samples: ① exclude samples with cancer when enrolled in 2013; ② exclude samples with variable missing numbers greater than 5 and samples lost to follow-up; ③ perform natural logarithm transformation on variables with right-skewed distribution; ④ exclude outlier samples according to the standard of deviation from the mean 8 times the interquartile range; ⑤ use multiple imputation (PMM) algorithm to fill in the missing values ​​of samples; ⑥ exclude samples that offset the training set.

[0107] Operation S3, based on the restricted cubic spline (RCS) method, identify clinical indicators with a U-shaped association with the risk of death; in the Cox proportional hazard regression model, correct the actual age and set the 10%, 50%, and 90% quantiles as nodes of the clinical indicators. This step corresponds to Figure 1 Step 2 of the biological age assessment method proposed in the present invention.

[0108] Operation S4, for the clinical indicators identified in step S3 that are U-shapedly associated with the risk of death, variable conversion is performed based on the clinical optimal value identified by the RCS curve. This step corresponds to Figure 1 Step 3 of the biological age assessment method proposed in the present invention. The formula for variable conversion based on clinical optimal values ​​is expressed as:

[0109]

[0110]

[0111] Among them, y i and Represent the original value and clinical optimal value of the i-th clinical indicator respectively. For clinical indicators with a U-shaped correlation with the risk of death, this transformation method can simultaneously convert values ​​above and below For clinical indicators that are linearly associated with the risk of death, no conversion is performed on the original values ​​of the clinical indicators.

[0112] Furthermore, this step resulted in 11 clinical indicators that were linearly associated with the risk of death, and 25 clinical indicators that were U-shaped associated with the risk of death. Therefore, there were a total of 11 original clinical indicator variables and 50 transformed variables. Among them, the RCS curves of the 11 clinical indicators that were U-shaped associated with the risk of death but had no significant linear effect were as follows: Figure 4 shown.

[0113] Operation S5, this step corresponds to Figure 1Step 4 of the biological age assessment method proposed in the present invention. In the training set, 33 predictors related to mortality risk were screened from the actual age and the 61 variables generated in step S4 using Cox regression based on LASSO. These included the original variables of actual age CA, systolic blood pressure SBP and red blood cell volume distribution width RDW, as well as 30 converted variables of 15 clinical indicators.

[0114] Furthermore, multivariate Cox regression was used to estimate the weight of each predictor, and the unit of the weight was the unit of the clinical indicator. In the multivariate Cox regression, the original measured values ​​of actual age CA, systolic blood pressure SBP and red blood cell volume distribution width RDW, as well as the transformed values ​​y of 15 clinical indicators were included. i1 and i2 This weight is used for PAI and its formula is expressed as:

[0115]

[0116] Among them, the death prediction factors identified and selected according to this embodiment and their estimated weights are shown in Table 1.

[0117] Operation S6, calculate PAI and ΔPAI on the test set samples and verify their predictive ability for the risk of death and new diseases. This step corresponds to Figure 1 Steps 5 and 6 of the biological age assessment method proposed in the present invention. The specific analysis process is as follows:

[0118] Analysis A1: In the test set samples, the clinical indicators with a U-shaped correlation with the risk of death were transformed according to the parameters pre-provided by the biological age assessment system provided by the present invention (Table 1).

[0119] Analysis A2, according to the formula predefined in the biological age assessment system provided by the present invention, calculate the PAI and ΔPAI of the sample. The relationship between PAI and ΔPAI is as follows: Figure 5 shown.

[0120] Analysis A3: In the test set, the Cox proportional hazards regression model was used to evaluate the discrimination (C index and time-dependent AUC index), calibration (slope of the calibration curve and outcome), and clinical decision-making value of PAI in predicting mortality risk. The results are shown in Table 2. Compared with the actual age CA and the existing biological age PhenoAge, PAI has better discrimination and calibration in predicting mortality risk.

[0121] Table 2

[0122]

[0123]

[0124] Furthermore, the C-index and time-dependent AUC index were used to evaluate whether the actual age and a single clinical indicator can improve the prediction effect of mortality risk. The following three Cox regression models were compared: ① including actual age as a covariate; ② including actual age and a single clinical biomarker as a covariate; ③ including actual age and ΔPAI as a covariate. Figure 6 As shown in the data, ΔPAI can improve the prediction effect of actual age and gender on the risk of death, and the prediction performance is better than any single clinical indicator.

[0125] Analysis A4: In the test set, Cox proportional hazards regression analysis was used to evaluate the predictive effect of ΔPAI on new cardiovascular disease. First, samples with abnormal electrocardiograms and existing cardiovascular disease were excluded. ΔPAI was subjected to standard normal transformation, and PAI was analyzed as a continuous variable and a categorical variable. When used as a categorical variable, the samples were divided into four groups using the PAI quartiles. Multivariate Cox regression was performed to compare the risk of cardiovascular disease and its subtypes in the three groups with higher PAI relative to the group with the lowest PAI, using cardiovascular disease, coronary heart disease, stroke, and acute coronary syndrome as outcomes. This analysis compared two models. The baseline model only included ΔPAI and actual age as covariates, and the full model additionally included 11 CVD risk factors as covariates, including sex, hypertension, hyperlipidemia, diabetes, obesity, smoking, drinking, diet, sleep duration, exercise frequency, and family history of cardiovascular disease.

[0126] The results are as follows Figure 7 As shown in the results, when ΔPAI was used as a continuous variable to predict the risk factors of incident cardiovascular disease and its subtypes, the relative risk ratios (HR per SD) of ΔPAI for incident cardiovascular disease, coronary heart disease, acute coronary syndrome, and stroke were 1.18 (95% confidence interval: 1.13, 1.24), 1.13 (1.07, 1.18), 1.22 (1.12, 1.34), and 1.40 (1.28, 1.52), respectively. When the samples were grouped by PAI quartiles, the risk of cardiovascular disease in the highest quartile of ΔPAI was much higher than that in the lowest quartile, especially stroke (HR = 2.35 [95% confidence interval: 1.76, 3.13]). When we further adjusted for 11 traditional risk factors, the effect size of ΔPAI was weakened, but it was still statistically significant in CVD, ACS, and stroke.

[0127] Furthermore, the biological age assessment system provided by the present invention is used to perform the second aspect of this embodiment. In this embodiment, PAI is estimated in the UK Biobank sample and its predictive value for death is evaluated, and operations S7 to S9 are performed. Among them, S7 corresponds to Figure 2The data collection module, nonlinear effect identification module and biological age assessment module of the provided biological age assessment system, operations S8 and S9 describe the application of the aging speed assessment system.

[0128] Operation S7, for example, to ensure the data quality of the UKB sample, first exclude samples that were lost to follow-up, had no blood drawn on the day of enrollment, had missing death time, had cancer at baseline, or had missing clinical indicators. Secondly, natural logarithm transformation was performed on variables with right-skewed distribution. According to the method described above, clinical indicators that were U-shapedly associated with the risk of death were transformed using the clinical optimal values ​​estimated from the training set of the "Dongfeng-Tongji" cohort (see Table 1 for details). Finally, the model weights (see Table 1 for details) were used to calculate the sex-specific PAI for 298,284 samples.

[0129] In operation S8, the Cox proportional hazards regression model was used to evaluate the discrimination (C index and time-dependent AUC index), calibration (slope of the calibration curve and outcome) and clinical decision-making value of PAI in predicting the risk of death. The results are shown in Table 3. Compared with the actual age CA and the existing biological age PhenoAge, the discrimination and calibration of PAI in predicting the risk of death were better.

[0130] Table 3

[0131]

[0132]

[0133] Furthermore, the C-index and time-dependent AUC index were used to evaluate whether the actual age and a single clinical indicator can improve the prediction effect of mortality risk. The following three Cox regression models were compared: ① including actual age as a covariate; ② including actual age and a single clinical biomarker as a covariate; ③ including actual age and ΔPAI as a covariate. Figure 8 As shown in the data, ΔPAI can improve the prediction effect of actual age and gender on the risk of death, and its prediction performance is better than any single clinical indicator.

[0134] Operation S9 evaluated the predictive effect of PAI on age-related diseases based on the UK Biobank. The predictive value of ΔPAI for the incidence of 9 age-related diseases was evaluated in the UK Biobank, including cancer, ischemic heart disease, stroke, chronic obstructive pulmonary disease, diabetes, dementia, chronic kidney disease, hearing loss and chronic liver disease. For each specific disease, cases that had been diagnosed before enrollment were excluded. Subsequently, a sex-stratified Cox model was used to adjust for age, center, smoking, drinking and Townsend Index. The results are shown in Fig. 9It was shown that ΔPAI was significantly correlated with all nine diseases, with the strongest correlation with chronic kidney disease (relative risk ratio HR = 1.59 [95% confidence interval: 1.58, 1.61]) and chronic liver disease (relative risk ratio HR = 1.48 [95% confidence interval: 1.45, 1.50]). It is worth noting that ΔPAI and ΔPhenoAge, as comprehensive aging indices, performed better than any single biomarker in predicting nine diseases except diabetes. Compared with ΔPhenoAge, ΔPAI performed better in predicting eight diseases except cancer, and had a significant improvement in predicting stroke and dementia.

[0135] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a biological age assessment system, characterized in that: The following steps are involved: (1) Collect clinical indicators of subjects; (2) performing nonlinear Cox proportional hazards regression analysis on the clinical indicators described in step (1) and the risk of death; (3) For the clinical indicators obtained in step (2) that are not nonlinearly associated with the risk of death, the original values ​​are not converted; for the clinical indicators obtained in step (2) that are U-shapedly associated with the risk of death, the original measured values ​​are converted based on their optimal values; the optimal value is defined as the value of the clinical variable in the restricted cubic spline curve when the relative risk of death is the lowest; (4) Predicting the risk of death based on the actual age, the original clinical indicators obtained in step (3), and the converted variables, using the LASSO algorithm for variable screening in the Cox proportional hazard regression model; (5) Calculate the biological age PAI of each sample based on the variables screened in step (4); (6) Based on the biological age PAI calculated in step (5) and the actual age of the sample, the biological aging rate ΔPAI of each sample is calculated.

2. The method for constructing a biological age assessment system according to claim 1, characterized in that: In step (1), the clinical indicators are height, weight, blood pressure, blood lipids, blood sugar, routine blood tests and blood biochemical tests, and the skewed distribution indicators in the clinical indicators are transformed into normal ones.

3. The method for constructing a biological age assessment system according to claim 1, characterized in that: In step (2), the model used for the nonlinear Cox proportional hazards regression analysis adopts the restricted cubic spline method, with actual age and a single clinical indicator as covariates, and the nodes of the clinical indicator are selected at the 10%, 50% and 90% quantiles.

4. The method for constructing a biological age assessment system according to claim 1, characterized in that: In step (3), the formula for variable conversion of the clinical indicator is: Among them, y i and represent the original value and optimal value of the i-th clinical indicator respectively.

5. The method for constructing a biological age assessment system according to claim 1, characterized in that: In step (4), the Cox proportional hazards regression model is expressed as: in, and h0(t) represent the hazard rate function and baseline hazard function at time point t, respectively, β i represents the regression effect value of the i-th indicator, y i· represents the i-th variable of all samples, and p is the number of clinical indicators. The covariates included in the Cox proportional hazard regression model include actual age, the original values ​​of the clinical indicators that are linearly associated with death, and the transformed values ​​of the clinical indicators that are nonlinearly associated with death.

6. The method for constructing a biological age assessment system according to claim 1, characterized in that: In step (5), the PAI calculation formula for sample j is: Where k is the number of variables selected for calculating PAI, y i· represents the value of the i-th variable of all samples, represents the weight of the i-th variable.

7. The method for constructing a biological age assessment system according to claim 5, characterized in that: Variable weights From the multiple linear regression analysis; in this multiple linear regression analysis, if a clinical indicator has a U-shaped association with the risk of death, only one variable transformation value is selected into the model by the LASSO algorithm, that is, y i or i2 is selected into the model by the LASSO algorithm, then y is retained at the same time i and i2 Used to calculate PAI.

8. The method for constructing a biological age assessment system according to claim 1, characterized in that: In step (6), the ΔPAI calculation formula for sample j is: Among them, x j represents the actual age of the jth sample, and represent the calculated slope and intercept terms in the linear regression of PAI on chronological age, respectively.

9. A biological age assessment system, characterized in that: include: Clinical index collection module: used to collect clinical indexes of subjects; Nonlinear effect identification module: Based on nonlinear Cox proportional hazard regression analysis, clinical indicators with a U-shaped correlation with the risk of death are identified; the value of the clinical variable when the relative risk of death in the restricted cubic spline curve is the lowest is defined as the clinical optimal value; Nonlinear effect conversion module: For clinical indicators that have a U-shaped correlation with the risk of death, their original measured values ​​are converted based on their optimal values; Variable screening module: Based on the candidate variables, a Cox proportional hazard regression model is constructed, and the LASSO algorithm is used for variable screening; Biological age estimation module: Calculates biological age PAI based on the predictors obtained by the nonlinear effect identification module and the variable screening module; Biological aging rate estimation module: Calculate the biological aging rate ΔPAI based on the biological age PAI and actual age.

10. A biological aging assessment method, characterized in that: The following indicators of the samples were collected: actual age (CA), in years; body mass index (BMI), in kg / m 2 , defined as weight divided by the square of height, weight is in kg, height is in m; Systolic blood pressure SBP, in mmHg; fasting blood glucose FBG, in mmol / L; total cholesterol TC, in mmol / L; red blood cell count RBC, in 10 12 / L; mean corpuscular volume MCV, in fL; mean corpuscular protein MCH, in pg; coefficient of variation of red blood cell volume distribution width RDW, in %; white blood cell count WBC, in 10 9 / L; Lymphocyte percentage LYM, unit is %; The unit of monocyte percentage MONO is %; the unit of alkaline phosphatase ALP is μ / L; the unit of aspartate aminotransferase AST is μ / L, the unit of alanine aminotransferase ALT is μ / L; the unit of direct bilirubin DBIL is μmol / L; the unit of serum creatinine SCR is μmol / L; the unit of urea nitrogen BUN is μmol / L; and the skewed distribution index is transformed by natural logarithm; the PAI and ΔPAI of the sample are calculated according to the biological aging assessment system, so as to assess the biological age and biological aging rate of the sample; When ΔPAI>0, it means that the biological aging rate exceeds that of peers; when ΔPAI=0, it means that the biological aging rate is equal to that of peers; when ΔPAI<0, it means that the biological aging rate is lower than that of peers.

11. Use of the method according to claim 10 in assessing the degree of aging and predicting the risk of death or disease.

Images