Chronological age prediction does not prove biological age measurement

In a computational preprint analyzing human data from NHANES and allogeneic haematopoietic stem-cell transplantation recipients, researchers evaluated how chronological-age prediction relates to biological age measurement. They demonstrated that training algorithms to predict chronological age identifies an age-task statistic rather than a true biological-age construct. Even optimal age prediction permits mutually incompatible biological interpretations. Squared-error loss penalizes score dispersion without defining its biological direction, meaning perfect age recovery eliminates age gaps even if underlying measurement heterogeneity remains. In transplant recipients, blood scores showed excess donor-lineage affiliation. In NHANES participants, age-trained scores improved five-year mortality predictions beyond chronological age and background. However, directly modeling source measurements or supervising directly for mortality achieved greater predictive gains.