Missing-data choices affect aging clocks in human blood
Computational tests on 4,798 blood samples found that replacing missing readings with zero produced stronger bias that varied with the size of aging-clock estimates.
Our read
Biomarkers- Studied:
- Computational study of blood samples from US adults, against other imputation methods
- What changed:
- Replacing missing readings with zero increased bias in aging-clock estimates compared with other imputation methods
- The catch:
- The analysis did not measure which clock best captured biological aging
- Why it’s aging research:
- It builds or tests a tool for aging research.
Editor of record: Boris Djordjevic · AgingResearch.News
In blood samples from US adults aged 33–43, researchers computationally tested how data processing affected epigenetic clocks. These tools estimate biological aging from DNA methylation, a pattern of chemical tags on DNA. They began with 4,798 samples and compared ways to replace readings removed by quality-control checks. Both types of check removed readings whose distribution differed from those retained.
Replacing missing readings with zero—effectively leaving those DNA sites out of each sample’s clock calculation—produced stronger proportional bias across all clocks tested. This means differences between methods varied with the size of the clock estimate. Other tested methods, including filling gaps with average or middle values, did not appear to show consistent patterns of bias. The authors recommended estimating missing values rather than leaving sites out.
Why it matters
Epigenetic clocks are widely used in aging research. The work addresses how researchers should handle missing measurements when constructing these estimates of biological aging.
Caveats
The analysis compared calculation methods rather than testing which clock best measured biological aging. It used one blood dataset from US adults aged 33–43.
- DNA methylation
- Epigenetic clock
- Distribution of removed versus retained DNA methylation beta values
- DNA methylation beta values
The paper
University of North Carolina
Epigenomics, Jul 20, 2026, Open access
Comments
Be the first to comment.