Guiding STAR
Navigate the S.T.A.R.s to discover your ideal epigenetic biomarker from 41 clocks and 1,693 proxies
Loading biomarker data across all domains...
Select the S.T.A.R. domains that matter for your research
S Stable - Robustness to Confounders
✓ Analyzing all stability data: biological ICC and technical ICC
T Treatment Response - Filter by Intervention Type
Preferred Effect Direction
Select which direction should rank higher:
A Associations - Filter by Disease Category
R Risk - Filter by Outcome Type
✨ Your Personalized Biomarker Rankings
Ranked by z-score performance across your selected S.T.A.R. domains
Calculating Z-Scores...
Analyzing biomarkers across your selected S.T.A.R. domains
🧮 How Your Biomarker Score is Calculated
S Stable Score Calculation
Stability Score: Pooled ICC from meta-analysis across control datasets (0-1 scale, higher = more robust to technical and biological confounders)
T Treatment Response Score Calculation
For each intervention response:
- Significance Score: 1.0 if p-value < 0.05, otherwise 0.3
- Effect Score: min(|effect size| ÷ 0.5, 1.0) to normalize large responses
- Response Score: (Significance + Effect) ÷ 2
Final Treatment Response Score: Average across all relevant intervention responses
A Associations Score Calculation
For each baseline disease association:
- Significance Score: 1.0 if p-value < 0.05, otherwise 0.3
- Effect Score: min(|beta coefficient| ÷ 2, 1.0) to normalize large effects
- Association Score: (Significance + Effect) ÷ 2
Final Associations Score: Average across all relevant disease associations
R Risk Score Calculation
For each future disease or mortality outcome:
- Significance Score: 1.0 if p-value < 0.05, otherwise 0.3
- Effect Score: min(|beta coefficient| ÷ 2, 1.0) to normalize large effects
- Risk Score: (Significance + Effect) ÷ 2
Final Risk Score: Average across all relevant future disease and mortality predictions
Your Personalized Biomarker Recommendations
STAR Score Distribution
Distribution of all biomarkers across selected domains (your top 10 highlighted)
Top 10 Recommended Biomarkers
Ranked by composite z-score across selected S.T.A.R. domains
Overview
Guiding Star uses evidence-based z-score calculations to rank biomarkers across the S.T.A.R. Framework. Each domain employs domain-specific statistical methods, then scores are normalized and combined into a composite ranking.
Stable Domain
Metric: Intraclass Correlation Coefficient (ICC)
avgICC = (biological_ICC + technical_ICC) / 2
z = (avgICC - mean) / SD
Higher ICC = more robust to technical and biological confounders
Treatment Response Domain
Metric: HRS population-scaled effect sizes
mean_response = mean(scaled_estimates)
SD_response = SD(scaled_estimates)
z = -mean_response / SD_response
Negative mean (rejuvenating effect) → positive z-score performance
Associations Domain
Metric: Meta-analysis z-scores (baseline disease links)
z_i = beta_i / SE_i
z_combined = Σ(z_i) / sqrt(n)
Stouffer's method aggregates evidence across filtered phenotypes
Risk Domain
Metric: Meta-analysis z-scores (future disease & mortality)
z_i = beta_i / SE_i
z_combined = Σ(z_i) / sqrt(n)
Includes Cox proportional hazards models for time-to-event analysis
Composite Score Calculation
- Calculate raw z-scores for each biomarker in each selected domain
- Rank-based normalization within each domain: sort by z-score, assign ranks 1-N, map to 0-1 scale (lowest z-score → 0, highest → 1)
- Filter by biomarker type: apply normalization separately for epigenetic clocks vs. proxies based on toggle selection
- Weighted averaging: STAR_score = Σ(rank_score × weight) / Σ(weights)
- Final ranking: sort by composite STAR score (higher = better overall performance)
User Filters
Users can customize biomarker selection by filtering specific phenotypes, intervention types, or outcome categories:
- Treatment Response: Filter by intervention type (lifestyle, pharmacological, supplement)
- Associations: Filter by disease category (cardiovascular, metabolic, neurological, cancer)
- Risk: Filter by outcome type (all-cause mortality, future disease)
📚 Learn More: For complete statistical methods and validation, see our published research and data usage guidelines.