The algorithm that learned to read the body, cell by cell

Osamu Hirose of Kanazawa University and Emanuele Rodolà of Sapienza University of Rome built a new kind of alignment tool for spatial transcriptomics — the maps that show where thousands of genes are switched on inside a tissue. Instead of squishing measurements into pixels, their Domain Elastic Transform matches cells using both position and gene activity, keeping single-cell detail intact. It beat rival methods across 90 test cases, handled maps of a million points using less than a gigabyte of memory, and even aligned mouse embryos one day of development apart. Published in IEEE Transactions on Pattern Analysis and Machine Intelligence.