REVIVE
Jung S, Hodar JA, Badam TVS, Del Sol A
2025 · Aging
Great efforts have been devoted to discovering rejuvenation strategies that counteract age-related functional decline and improve cellular functions in humans.
Abstract
From the original paper, Aging · PubMed
Great efforts have been devoted to discovering rejuvenation strategies that counteract age-related functional decline and improve cellular functions in humans. However, new discoveries are currently driven by expert knowledge and require large amounts of resources. Here, we present REVIVE (Rejuvenation Estimation Via Insightful Virtual Experiments), the first computational framework for systematically predicting chemical and genetic perturbations that can restore a youthful transcriptional state based on gene expression data. REVIVE leverages age predictions to detect significant rejuvenating effects and quantifies the impact of perturbations on the hallmarks of aging. When applied to a large-scale in silico screen of more than 10000 compounds and genetic perturbations, REVIVE recapitulates known interventions as well as 477 novel compounds that restore a more youthful transcriptional state improving multiple aging hallmarks. Finally, we demonstrate the utility of REVIVE for repurposing perturbations to revert aged transcriptional states.
Summary
Level-2 proxy inverse design for rejuvenation: a GLM-LASSO multi-tissue transcriptional aging clock (trained on MultiTIMER RNA-seq with GTEx young/old DEG filtering) scores LINCS L1000 chemical and genetic perturbation profiles (GAN pseudo-RNA-seq) for predicted age reduction and Aging Atlas hallmark rescue; a repurposing mode ranks perturbations against AgeAnno young/old DEG signatures. Validation is entirely in silico (clock CV, 80-condition gold standard, DrugAge concordance, large LINCS screen); the paper notes metformin generally did not significantly reduce predicted cellular age.
Why this class
Level 2 because the method instantiates the inverse-design objective through an empirical or learned proxy response map rather than a mechanistic intervention-dependent model. Representation family is signature / state-matching. Cited in §3.3 of the review. Editorial rationale pending review by the maintainer.