Tumor tissues exhibit high heterogeneity, characterized by diverse cell types, disordered gene expression, and indistinct regional boundaries, making it difficult for traditional spatial transcriptomics techniques to fully reveal their characteristics. Conventional methods often treat sampling points as independent "houses", analyzing cell types and gene expression at each point separately, which can lead to errors.

To address this challenge, Professor Ye Kai's team at Xi'an Jiaotong University (XJTU) proposed a computational framework based on the spatial distribution of gene expression. Their research, titled "STMiner: gene-centric spatial transcriptomics for deciphering tumor tissues", was published in Cell Genomics on Feb 12.

Professor Ye's team has long been dedicated to developing bioinformatics methods and applying them to the early diagnosis and precision treatment of genetic diseases and tumors.

The new method, known as STMiner, utilizes a two-dimensional Gaussian mixture model and optimal transport theory to transform discrete sampling points into a continuous gene distribution map – akin to drawing a complete "resident distribution map" from scattered "household information".

This method provides a more accurate representation of tumor tissue, reduces interference, and enables scientists to better observe gene spatial distribution patterns. Using STMiner, the research team successfully identified distinct gene expression patterns in complex, poorly defined tumor regions, precisely delineated the boundaries between normal and abnormal tissues, and uncovered previously overlooked key genes and spatial structures. These discoveries open new avenues for tumor research and are expected to provide critical insights for future precision therapies.

This work represents another major achievement from the interdisciplinary collaboration at XJTU's Center for Information and Biomedical Research. 

The first author of the paper is Sun Peisen, a PhD student at the School of Automation, Faculty of Electronic and Information Engineering, and Ye is the corresponding author.

Schematic of STMiner computational strategy.