Journal: bioRxiv
Article Title: Deep learning predicts haematopoietic stem cell ageing from 3D chromatin images
doi: 10.64898/2025.12.11.693143
Figure Lengend Snippet: a. Schematic representation of the inference pipeline using ChromAgeNet, where the two different mechanism of actions of the epigenetic drugs are depicted: Inhibitors of Rho GTPase (CASIN and RhoA inhibitor) and modulators of H3K9 methylation (UNC0646 and IOX1). Designed with BioRender. b. Distribution plots showing soft voting-aggregated and calibrated ChromAgeNet scores at nucleus level for young and aged HSC, along with different treatments of aged HSCs. Probability values near 0 reflect more aged-like phenotypes, while values near 1 reflect more young-like phenotypes. c. Heatmap showing normalized values of top selected SHAP features (columns) distributed over images from different young, aged and drug-treated aged HSCs (rows), all acquired by the same microscopist and microscope.
Article Snippet: Where indicated, cells were treated with 100 μ M Rhosin (RhoAi) [ , ], 5 μ M CASIN [ ], 50 μ M IOX1 (8-hydroxyquinoline-5-carboxylic acid), which mimics 2-OG ( α -KG) and blocks the catalytic activity of lysine demethylases (from Tocris Biotechne), 0.25 μ M UNC0646 (Sigma), a well-described selective G9a/GLP methyltransferase inhibitor [ ] or left untreated.
Techniques: Methylation, Microscopy