PULSE AI reconstructs molecular profiles from routine clinical data
bravo_abad · x · 2026-08-27
Wei Wu et al. introduce PULSE, a self-supervised framework using longitudinal multimodal alignment to reconstruct expensive proteomic and metabolomic biomarkers from routine clinical measurements (labs, history, retinal images). It models patient visits as trajectories through a physiological state space. On UK Biobank data, PULSE reconstructed 251 metabolomic biomarkers.
More from Research
- Visual General Intelligence White Paper Published on arXiv — HirokatuKataoka · 2026-08-27
- OpenMythos: open-source theoretical reconstruction of Claude's architecture hits 14.8k stars — Shruti_0810 · 2026-08-27
- Community open-sources theoretical reconstruction of Claude Mythos — Shruti_0810 · 2026-08-27
- Paper Fixes 2D-3D Fluid Coupling Mismatch with Nonlinear Wave Model — ssh4net · 2026-08-27
- MIT researchers discover method to enable real logical reasoning in LLMs — mdancho84 · 2026-08-27
- ArayaGlobal emphasizes quality over quantity in open data contributions — kaixhin · 2026-08-27