Predictive adequacy and measurement dependence of metabolic representations in controlled synthetic systems

In 20 synthetic confirmation cohorts (5,115 individuals each) predicting year-25 outcomes from years 1–20, only five adequacy groups met all requirements, while none of the twelve Phi groups did.

• Why it matters: Metabolic representations can look compact or change repeatably without being adequate, so evaluation needs explicit reference information, measurement conditions, and definitions of change.
• What they did: Six stipulated stochastic scenarios generated cohorts; three representation families (Phi, a larger measurement/history collection, and an outcome-guided learned embedding) were tested against complete longitudinal history with 42 adequacy groups and 2,850 simultaneous upper bounds.
• The result: Of 393 failed adequacy predicates, 321 had estimates within the margin but upper bounds above it, and measurement dependence was strong (e.g., the panel embedding used eight coordinates but depended on 72 upstream engineered values).