MSU researchers highlight technological improvements in identifying gene traits
Two papers by Michigan State University researchers in spatial transcriptomics were recently published in Nature Communications. The technology presented has the potential to make an impact in cancer treatments, as genetic information about the environment surrounding tumors can make an impact on a patient’s immune response.
Spatial transcriptomics is a molecular biology tool that allows researchers to see which genes are active in different parts of a tissue—and where that activity occurs. Like using a map program with gene sequencing, scientists can see the location and biological activity of genes at the same time. Researchers at MSU have focused on improving the technology usage in the field in their latest studies.
The papers, “STANCE: a unified statistical model to detect cell-type-specific spatially variable genes in spatial transcriptomics” and “Rotation-invariance is essential for accurate detection of spatially variable genes in spatial transcriptomics,” were co-authored by members of MSU’s Department of Statistics and Probability. Yuehua Cui, professor, and Haohao Su, graduate student, contributed to both studies, while Yuesong Wu, graduate student, and Bin Chen, associate professor in the Department of Pharmacy and Toxicology, were co-authors for the STANCE publication.
The development of STANCE was supported by a Strategic Partnership Grant from the MSU Office of Research and Innovation and the High-Performance Computing Center (HPCC).
STANCE addresses major challenge
One of the major challenges in spatial transcriptomics is finding spatially variable genes (SVGs). These genes are either turned on or off in specific spatial regions instead of randomly across the tissue. They are also active in certain cell types, called cell-type-specific spatially variable genes (ctSVGs). Locating these genes are challenging because gene patterns can overlap or change depending on how the tissue is viewed or rotated.
To address this, a quartet of MSU researchers introduced a statistical method called STANCE (Spatial Transcriptomics ANalysis of genes with Cell-type-specific Expression). STANCE uses statistical methods to study SVGs more accurately by combining data on gene activity, cell type compositions and tissue location. The method works regardless of how the tissue is oriented through a two-step process: 1) finding genes that vary by location; and 2) testing genes that are specific to certain cell types.
STANCE not only gets consistent results, but it detects both SVGs and ctSVGs. Earlier tools were limited in finding spatial variation, but unable to determine the cell types involved. STANCE also allows researchers to see how much each cell type contributes to the spatial pattern for a given gene. This gives biological insight about which cells are important in different parts of tissue.
Understanding gene expressions within cell types across a tissue section is one of the major impacts of STANCE, according to Cui.
“In many diseases, such as cancer, the spatial location of cells matters,” Cui said. “Cells near blood vessels behave differently, immune cells may cluster in certain regions, etc. Identifying which genes are changing in space, especially within specific cell types, can reveal mechanisms of disease and help find diagnostic markers or targets for therapy.”
Insights from spatial patterns has the potential to help in diagnostics or designing treatments, Cui said. This is particularly applicable in cancer, where the environment around tumors matters for immune response.
The researchers are working on a new, faster method to find both SVGs and ctSVGs. It will work like STANCE but will be designed to handle larger datasets more efficiently, making it easier to use for everyday spatial transcriptomics research.
Rotation-sensitive methods can lead to unreliable results
Another benefit of STANCE is how it addresses methods that are rotation sensitive. Many current methods for finding SVGs give different results depending on how a tissue is positioned on a slide.
In Cui and Su’s most recent paper, the duo identified the technical pitfalls of the current methods and discussed strategies for rotation-invariant methods, which enhances the detection of SVGs.
“Without rotation invariance, you might draw the wrong conclusions about which genes are spatially active,” Cui said. “That could lead to mistaken biological inferences, especially in disease studies or comparing multiple tissue slices.”
Reliability and reproduction of results is improved with rotation invariance, according to Cui. “When people across different labs prepare tissue differently, or align slices differently, you want your analysis tool to give results that don’t depend on those arbitrary choices,” he said.
The study also lays the groundwork for future scientists to evaluate and build better tools. The researchers also suggest that existing datasets need re-examination under methods that ensure rotation invariance to confirm previous findings.
“This is a methodological foundation piece,” Cui said. “It doesn’t point to a new gene or disease mechanism but improves how we can trust spatial transcriptomics as a technology.”
- Categories: