Associations Dashboard

Evaluate epigenetic clock associations with baseline diseases and phenotypes using random-effects meta-analysis

Datasets -- Individuals -- Samples --

Desktop Required

For analysis features, please use the desktop version of this page. The interactive dashboard requires a larger screen.

Step 1: Select one Disease or Phenotype!

Step 2: Select multiple Clocks or Biomarkers!

Current Selection

Selected Disease/Phenotype:

No phenotype selected

Selected Clocks: 0

No clocks selected

Select a phenotype to see contributing datasets

Instructions:

1. Select ONE disease or phenotype from the left panel
2. Select multiple epigenetic clocks to analyze
3. Use the tabs above to explore different visualizations
4. Download results from the Data Table tab

Use Case 1: Biomarker Discovery Workflow

Step 1 Toggle to "Baseline" and select a disease of interest

Step 2 Use "Select all" or choose specific clock categories

Step 3 View Associations tab for z-score rankings

Step 4 Download data from Data Table tab for further analysis

Use Case 2: Group Comparison Analysis

Step 1 Select a baseline phenotype (disease or continuous trait)

Step 2 Choose clocks of interest from categorized lists

Step 3 View Comparison tab - automatically shows binary or quartile analysis

Step 4 Examine group means, error bars, and contributing datasets

Prognostic Analysis Methodology

Understanding our approach to meta-analysis of epigenetic clock associations with disease outcomes

Analysis Pipeline Overview

Individual Studies

Fit models in each of 180+ datasets using standardized covariates

Effect Aggregation

Combine z-scores across studies with similar phenotypes

Ranking & Visualization

Generate heatmaps and rankings of biomarker associations

Flexible Regression Framework

Model Selection

Appropriate models are automatically selected based on outcome type:

Binary Outcomes

glm(outcome ~ scale(clock) + cAGE + cFEMALE, family=binomial)

Continuous Outcomes

lm(outcome ~ scale(clock) + cAGE + cFEMALE)

Ordinal Outcomes

polr(outcome ~ scale(clock) + cAGE + cFEMALE)

Survival Outcomes

coxph(Surv(time, event) ~ scale(clock) + cAGE + cFEMALE)

Meta-Analysis Framework

Random-effects meta-analysis using the metafor package:

β_i = θ + u_i + ε_i

β_i: Effect size from study i
θ: Pooled effect size
u_i ~ N(0, τ²): Between-study heterogeneity
ε_i ~ N(0, SE_i²): Within-study sampling error

rma(yi = beta, sei = se, method = "REML")

Standard Covariates

scale(Clock): Standardized epigenetic clock (mean=0, SD=1)
cAGE: Chronological age
cFEMALE: Sex (0=male, 1=female)

Extracted Parameters

From each fitted model, we extract:

β: Regression coefficient for standardized clock
SE: Standard error (SE = |β/z|)
z: Test statistic (z-score or t-statistic)
Model Type: Logistic, Linear, Ordinal, or Cox

Meta-Analysis Implementation

For clock-phenotype combinations:

Multiple Studies: Random-effects meta-analysis via REML
Single Study: Original values retained
Failed Models: Silently skipped (error handling)
Output: Pooled β, SE, z-score, and p-value

Key Features

• 180,000+ associations analyzed
• >100 disease-related phenotypes
• Standardized effect size reporting
• Interactive filtering and sorting
• Dataset-level transparency

Applications

• Biomarker discovery for specific diseases
• Clock performance comparison
• Clinical translation planning
• Study design optimization
• Prioritization of validation studies

Interpretation Guide

Positive Z-Score

Higher epigenetic age associated with increased disease risk

Negative Z-Score

Lower epigenetic age associated with decreased disease risk

|Z| ≥ 1.96

Statistically significant association (p < 0.05)

← Back to Front Page

PowerBI Version Responsive Dashboard Stability Dashboard