Research outcomes
Building Spatial Gene Expression Maps Without a Blueprint
— A Machine Learning Approach for Predicting Mutant Spatial Transcriptomes Without Training Data —
Building Spatial Gene Expression Maps Without a Blueprint
— A Machine Learning Approach for Predicting Mutant Spatial Transcriptomes Without Training Data —
A collaborative research team led by Kosuke Okouchi, Naoki Honda, Takaaki Matsui, and Takeshi Kondo has developed “ZENomix,” a novel machine learning method that predicts spatial transcriptomes from single-cell RNA sequencing data without requiring training data from mutant or diseased tissues.
The study demonstrated that ZENomix can accurately reconstruct gene expression maps in mutant zebrafish embryos and identify previously unknown genes regulated during early development. The predicted spatial expression patterns were experimentally validated by in situ hybridization analysis.
Because ZENomix can infer spatial gene expression patterns using only spatial data from normal tissues, the method is expected to accelerate developmental biology and disease research by enabling spatial interpretation of the rapidly growing amount of single-cell RNA-seq data worldwide.
The findings were published in the international journal Patterns on June 12, 2026.
DOI: 10.1016/j.patter.2026.101521
https://www.med.nagoya-u.ac.jp/medical_E/research/pdf/Pat_260612en.pdf
Laboratory of Biosystem Dynamics
https://bsw3.naist.jp/eng/courses/courses217.html
https://sites.google.com/view/matsui-labnaist/
( June 15, 2026 )
