Twin Attention AI Tracks Cells in Developing Worm Embryos
Published in Artificial Intelligence, Bioinformatics, Biology, Journals.
Publication date: October 10, 2026.
Memorial Sloan Kettering Cancer Center highlighted Twin Attention in its October 9 research update. The AI system identifies and follows individual cells during embryonic development, potentially reducing the manual work needed for large experimental screens. MSK links the research to Cell Reports Methods.
Twin Attention compares pairs of three-dimensional embryo images. It represents cells using both their location and relationships with neighboring cells, helping distinguish them as the tissue changes.
In experiments with the roundworm C. elegans, MSK reports cell-identification accuracy of approximately 93%–97% as embryos expanded to more than 500 cells. The team also analyzed 700 embryos with genes silenced, completing work in hours that would otherwise require hundreds of hours of manual analysis. The screen identified 29 genes whose silencing delayed gastrulation, a developmental stage involving inward cell movement.
This is a research method demonstrated in worm embryos. The reported accuracy does not establish performance in human tissues or clinical diagnosis. Its relevance lies in studying how cells organize and how gene perturbations disrupt development.
Primary source: MSK research update, October 9, 2026. Linked research: Cell Reports Methods.