Rejuvenation readouts cannot distinguish younger from less old

Treatments said to rejuvenate cells are usually judged by a single score that falls when cells look younger. But such a score shows only which way cells moved, not where they ended up: it also falls when cells simply stop looking old, for example by drifting toward a different cell type, without becoming any more like young cells. We ask a more direct question: after treatment, are the cells closer to young cells of the same type than they were before? In skin fibroblasts from a 96-year-old donor that were partially reprogrammed (Lu et al.), the published aging score fell well below the level of a 22-year-old donor's untreated cells, which looks like strong rejuvenation. Yet the same cells ended about twice as far from the young donor's cells as they started, even when judged only by the genes that make up that score. The same was true with other published aging genes, with all genes, with hundreds of randomly chosen gene sets, and with an age score we built that follows donor age across several independent cohorts. Our measure does register cells getting closer when young cells are mixed into old ones, so it can see real rejuvenation when it happens, and the result is not the two donors' cells simply becoming more alike. A falling aging score therefore cannot tell rejuvenation apart from cells moving away from their old state in some other direction. We provide a frozen age score, young reference points and code for measuring distance to young cells instead.