Search

Search Results

1 results found. Results per page: [ 20 ][ 40 ][ 60 ][ 80 ][ 100 ][ 200 ][ 300 ]

Sort by: [ Publication Date ][ Score ]

Year of publication: [ 2025 ][ 2024 ][ 2023 ][ 2022 ][ 2021 ][ 2020 ][ 2019 ][ 2018 ][ 2017 ][ 2016 ][ 2015 ][ 2014 ][ 2013 ][ 2012 ][ 2011 ][ 2010 ][ 2009 ][ Any ]

Direction: [ Desc ][ Asc ]

• Research Paper Volume 11, Issue 5 pp 1404-1426

  Age-dependent co-dependency structure of biomarkers in the general population of the United States

  Relevance score: 8.051737

  Alan Le Goallec, Chirag J. Patel

  Keywords: aging, biomarkers, correlations, machine learning, demographics

  Published in Aging on February 28, 2019

  Show abstract
  Hide abstract

  Phenotypic biomarkers (e.g. cholesterol, weight, and glucose) are important to diagnose and treat diseases associated with aging. However, while many biomarkers are co-dependent (e.g. glycohemoglobin and glucose), it is generally unknown how age influences their co-dependence. In the following, we analyzed 50 biomarkers in 27,508 National Health and Nutrition Examination Survey (NHANES) participants (age range: 20 to 80, mean age: 46.3 years old, sexes: 48.9% males, 51.1% females, ethnicities: 46.0% Whites, 27.8% Hispanics, 20.0% non-Hispanic Blacks, 6.1% others) to investigate how the co-dependency structure of common biomarkers evolves with age and whether differences exist between sexes and ethnicities. First, we associated the change in correlations between biomarkers with chronological age. We identified six trends and replicated our top finding (height vs. systolic blood pressure) in participants of the UK Biobank (N=470,895). We found that, on average, correlations tend to decrease with age. Secondly, we examined how biomarkers predict other biomarkers in participants of different age groups. We found 17 (34%) biomarkers whose predictability decreases with age and 5 (10%) biomarkers whose predictability increases with age. A limitation of this study is that it cannot distinguish between biological changes related to aging and generational effects. Our results can be interactively explored here: http://apps.chiragjpgroup.org/Aging_Biomarkers_Co-Dependencies/.

  

  Flowchart to analyze the age trajectories of the correlations.

  
  
  

  

  Changes in biomarkers with age. (A) Systolic blood pressure. Example of a biomarker that increases with age. (B) Upper leg length. Example of a biomarker that significantly decreases with age. (See discussion for explanation about the generational effect). (C) Volcano plot of the significance of the changes associated with age for the 50 biomarkers. The black horizontal line corresponds to the 0.05 significant threshold, after Bonferroni correction.

  
  
  

  

  Baseline correlation structure of the biomarkers. (A) Correlation matrix of the biomarkers.The biomarkers are ordered based on their clustering. Red is associated with a positive correlation, blue with a negative correlation. (B) Hierarchical clustering of the biomarkers. For each cluster, the number in green is the bootstrap probability (BP)--the percentage of bootstraps in which the cluster was present. The number in red is called the approximated-unbiased p-value (AU). AU is a better estimation of the unbiased p-value than BP, and the red boxes circle the significant clusters, based on this criterion, with alpha=0.95 (a cluster is marked as significant if its AU is greater than 95). The number in grey is the rank of the cluster, low numbers means the clustering happened early in the process. The height is the measure of the proximity between the two clusters being merged. The height is one minus the mean correlation between the two clusters, so two perfectly correlated biomarkers/clusters cluster at height zero, and two perfectly uncorrelated biomarkers/clusters cluster at height one. The first column compares males and females, the second column compares non-Hispanic Whites and Hispanics, the third column compares non-Hispanic Whites and non-Hispanic Blacks, and the fourth column compares the two control groups.

  
  
  

  

  Volcano plot of correlations changes with age. The black horizontal line corresponds to the Bonferroni-corrected significance threshold (0.05). The green dots are correlations that decrease with age but remain positive. The blue dots are correlations that decrease with age, starting positive and ending negative. The purple dots are correlations that decrease with age and remain negative. The yellow plots are correlations that increase with age but remain negative. The red dots are correlations that increase with age, starting negative and ending positive. The orange dots are correlations that increase with age and remain positive.

  
  
  

  

  Examples of different trajectories for changes in correlations with age. (A) Hematocrit vs. total calcium. Example of a correlation that decreases with age but remains positive. (B) Standing height vs. systolic blood pressure. Example of a correlation that decreases with age, starts positive and ends negative. In red, we replicated the analysis on the UK Biobank cohort. (C) Percentage of segmented neutrophils vs. number of lymphocytes. Example of a correlation that decreases with age and remains negative. (D) Hematocrit vs. chloride. Example of a correlation that increases with age but remains negative. (E) Percentage of segmented neutrophils vs. creatinine. Example of a correlation that increases with age, starts negative and ends positive. (F) Chloride vs. sodium. Example of a correlation that increases with age and remains positive.

  
  
  

  

  Correlations: Change of the means of the absolute values of the correlations with age. (A) Change of the mean of the absolute value of all the correlations with age. (B) Change of the mean of the absolute value of the correlations with age, limiting the analysis to correlations which significantly change with age.

  
  
  

  

  Correlations: Baseline differences between demographics. (A) Heatmap visualization of correlation structure. The upper left triangle shows the correlation structure for the first group of the comparison (e.g. males) while the lower right triangle of the matrix shows the difference between the second group and the first group of the comparison (e.g. difference between females and males). (B) Correlation between the 1,225 correlations of the first group (e.g. males) and the 1,225 correlations of the second group (e.g. females). The diagonal black line represents a perfect correlation. The further away from this line the points lie, the bigger the difference between the two groups, and the lower the correlation coefficient. (C) Volcano plots of the 1,225 differences in correlations between the groups. The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values. The vertical axis is not shared between the plots.

  
  
  

  

  Correlations: Differences in age trajectories between demographics. (A) Correlations between the rates at which the correlations are changing in the two compared groups, only taking into account the correlations for which significant changes were detected in both compared groups (e.g. males and females). (B) Volcano plots showing differences in the rates of change of correlations with age between different demographic groups. The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values.

  
  
  

  

  Predictabilities: Baseline predictabilities for the 50 biomarkers and significance of their changes with age. (A) Histogram of the predictabilities.Histogram of the predictabilities (R²) obtained on the 50 biomarkers using the full 27,508 samples. (B) Volcano plot of the changes in predictabilities with age.The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values.

  
  
  

  

  Changes of the predictabilities (R²) with age. (A) Predictabilities: Examples of different age trajectories. (A1) Albumin. Example of a biomarker whose predictability decreases with age. (A2) Glucose, serum. Example of a biomarker whose predictability increases with age. (B) Change of the means of the predictabilities with age. (B1) Change of the mean of all the predictabilities with age. (B2) Change of the mean of the predictabilities with age, limiting the analysis to biomarkers whose predictability significantly change with age.

  
  
  

  

  Predictabilities: Differences between demographics. (A) Correlation between the predictabilities of the 50 biomarkers in the first group and the predictabilities of the 50 biomarkers in the second group. The diagonal black line represents a perfect correlation. The further away from this line the points lie, the bigger the difference between the two groups, and the lower the correlation. (B) Volcano plots reporting the significance and the size of the baseline (calculated on the full age range 20-80) differences between the groups. The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values. (C) Volcano plots showing which predictabilities are changing at significantly different rates with age in different demographic groups. The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values.

  
  
  

  

  Regression coefficients: Baseline regression coefficients for the prediction of the 50 biomarkers and significance of their changes with age. (A) Heatmap of the regression coefficients on the full cohort. Each column corresponds to the 49 coefficients used to predict the values for one of the 50 biomarkers. (B) Volcano plot of the changes of the regression coefficients with age.The black horizontal line corresponds to the 0.05 significant threshold, after Bonferroni correction. The green dots are coefficients that decrease with age but remain positive. The blue dots are coefficients that decrease with age, starting positive and ending negative. The purple dots are coefficients that decrease with age and remain negative. The yellow plots are coefficients that increase with age but remain negative. The red dots are coefficients that increase with age, starting negative and ending positive. The orange dots are coefficients that increase with age and remain positive.

  
  
  

  

  Examples of different trajectories for changes in regression coefficients with age. (A) Target: Blood urea nitrogen. Predictor: Osmolality. Example of a coefficient that decreases with age but remains positive. (B) Target: Sodium. Predictor: Creatinine. Example of a coefficient that decreases with age, starts positive and ends negative. (C)Target: Sodium. Predictor: Blood urea nitrogen. Example of a coefficient that decreases with age and remains negative. (D) Target: Blood urea nitrogen. Predictor: Sodium. Example of a coefficient that increases with age but remains negative. (E) Target: Albumin, urine. Predictor: Creatinine. Example of a coefficient that increases with age, starts negative and ends positive. (F) Target: Osmolality. Predictor: Blood urea nitrogen. Example of a coefficient that increases with age and remains positive.

  
  
  

  

  Regression coefficients: Differences between demographics. (A) Heatmaps displaying the baseline differences in the regression coefficients between demographic groups. (B) Correlation between the 2,450 regression coefficients in the first group and the 2,450 regression coefficients in the second group. The diagonal black line represents a perfect correlation. The further away from this line the points lie, the bigger the difference between the two groups is, and the lower the correlation coefficient is. (C) Volcano plots reporting the baseline (calculated on the full age range 20-80) differences in regression coefficients between the groups. The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values. (D) Volcano plots showing which regression coefficients are changing at significantly different rates with age in different demographic groups. The horizontal black lines represent the threshold of significance of 0.05 for the Bonferroni corrected p-values.