Researchers in China have constructed a comprehensive three-dimensional spatial cell atlas tracking rice across its entire life cycle, from seed germination to flowering and grain filling. Released on October 6, 2026, and published in the international journal Cell, the milestone mapping effort provides a high-resolution window into the molecular architecture of the world’s vital staple crop.
Mapping the Super-Factory of Rice Growth
A single rice grain transforms through a complex, highly coordinated biological process where tens of thousands of cells divide, migrate, and specialize. Until now, observing this cellular choreography in complete detail across a full developmental timeline remained a major hurdle for agricultural scientists.
Using the japonica rice variety “Zhonghua 11” as their primary model, the research team integrated BGI’s proprietary Stereo-seq spatial omics technology, the DNBelab C4 single-cell library construction platform, high-throughput T7 gene sequencers, and artificial intelligence models alongside a high-quality T2T genome. The team successfully compiled data covering 10 major organ or tissue categories and 61 distinct developmental stages. The resulting dataset encompasses over 850,000 high-quality nuclear profiles and more than 340,000 spatial data units, successfully annotating 119 cell types and 133 cell subtypes.
The atlas captures not only the tissue distribution of various cell types but also records functional gene expression shifts across every stage of growth.
Decoding the Inner Workings of the Endosperm
Beyond mapping gross morphology, the atlas uncovers intricate biochemical specializations within the rice endosperm—the primary tissue determining grain nutritional profile and texture. Researchers discovered that developing endosperms operate with precise regional divisions. The dorsal peripheral region is heavily enriched with genes linked to carbohydrate metabolism and starch synthesis, while the ventral peripheral region favors storage protein synthesis.
Further mutational experiments demonstrated that altering these specific genes shifts the spatial accumulation patterns of starch and protein across the dorsal-ventral axis. Additionally, the team identified that key regulatory factors, such as OsARF1, exhibit variable transcriptional activity across different spatial environments and developmental phases, shedding light on complex pleiotropic genetic traits where a single gene influences multiple phenotypic outcomes.
An AI Model and Public Resource for Global Breeding
To maximize the utility of these findings for the broader scientific community, the consortium established an interactive online atlas supporting gene queries, spatial expression displays, and comparative analysis. Furthermore, the team developed the rice single-cell foundational model, RICE scGPT, serving as an artificial intelligence tool trained on single-cell datasets to assist in cell type annotation and dataset integration.
This achievement builds upon a 26-year collaborative legacy initiated in 2000, when Yuan Longping partnered with BGI to launch foundational rice genome sequencing efforts, culminating in the 2002 publication of the rice indica genome framework in Science. Today, scientific observation has moved past raw genetic sequences into spatial, cellular, and developmental dimensions, offering modern breeders unprecedented precision for targeted crop improvement.
