How CONCERT works: GP kernels model how far perturbation effects spread
marinkazitnik · x · 2026-10-25
Technical follow-up in the CONCERT release thread. Key insight: a gene knockout starts in perturbed cells and extends into the neighborhood — the same knockout can trigger different responses in a tumor core vs. a tissue border.
CONCERT learns a spatial kernel per perturbation with a Gaussian process variational autoencoder, describing how far an effect extends from perturbed cells and how it depends on the surrounding niche. With a trained model you can swap the perturbation, move a perturbed patch into another niche, or extend time after perturbation, on 2D slides or across 3D sections.
Related event: Harvard's CONCERT Predicts Spatial Spread of Gene Perturbations in Tissue(4 posts)→
More from Research
- New soft actuator weighs 15g, lifts 400g, costs under $3 to build — IanPritchard · 2026-10-25
- Three different arXiv papers are all named LeWAM, and researchers are begging for mercy — burny_tech · 2026-10-25
- Tinker's StableReinforce reverse-engineered: it's REINFORCE plus score centering — andersonbcdefg · 2026-10-25
- Moderna's phase 3 cancer vaccine cuts tumor recurrence by 42% — mariolefebvre · 2026-10-25
- NVIDIA Open-Sources 2,575 SimReady 3D Assets, 472GB With Physics Data — tinyfool · 2026-10-25
- New paper uses 288M job postings to measure where AI raises the value of work it can't do — soumitrashukla9 · 2026-10-25