Scientists at NYU have developed a nuclear morphometric pipeline (NMP) that employs unsupervised machine learning to analyze changes in nuclear size, shape, intensity, and foci. By clustering these features, the NMP accurately identifies bona fide and pre-senescent cell states in vitro and in vivo across muscle regeneration and osteoarthritis models, offering a standardized, high-throughput approach for senescence mapping in aging and disease contexts.
Key points
- Pipeline quantifies DAPI-stained nuclear size, circularity, intensity, and dense foci, then applies k-means clustering and UMAP to classify cell states.
- Validated across oxidative and genotoxic inducers (H₂O₂, etoposide, doxorubicin) and cell types (C2C12, 3T3-L1, primary FAPs, SCs, ECs, chondrocytes) by Ki67, γH2AX, SA-β-gal, and senolytic assays.
- In vivo mapping in young, aged, and geriatric mouse muscle and cartilage reveals dynamic, age-dependent distributions of senescent cell populations relevant to regeneration and osteoarthritis.
Why it matters: This standardized, ML-based approach transforms senescent cell detection, enabling scalable mapping of aging processes and targeted therapeutic interventions across tissues.
Q&A
- What is cellular senescence?
- How does nuclear morphology reflect senescence?
- What is UMAP and why is it used here?
- Why use unsupervised clustering instead of supervised learning?