Background

Spatial transcriptomics (ST) technology, which simultaneously analyzes the location information and gene expression levels of individual cells constituting a biological tissue, is revolutionizing the field of life sciences. Previous studies typically involved observing two-dimensional (2D) sections of tissues. However, actual organs and tumors possess complex three-dimensional (3D) structures. There is a growing consensus that 3D observation is essential to fully understand the intricate interactions between cells and the dynamic biological phenomena.

Recently, with the development of continuous section reconstruction techniques and high-resolution imaging methods, 3D ST data is rapidly accumulating. The problem is that the increasing size and complexity of the data are causing significant bottlenecks in the analysis and visualization processes. It has been difficult to place gene map data of millions of cells in 3D space and allow analysts to control it in real-time using existing technologies. Most existing tools suffer from slow processing speeds, often causing the screen to freeze, and the data analysis environment, such as Jupyter Notebook, is disconnected from the visualization program, interrupting the analysis flow.

Key Findings

The research team led by Professor Jian Yang at Westlake University in China has developed SpatialVista, a next-generation integrated visualization platform that can explore large-scale 3D ST data without delay, to solve this problem. The system they designed is characterized by its adoption of Web Graphics Library (WebGL) technology to maximize Graphics Processing Unit (GPU) acceleration. Thanks to this technology, they have successfully rendered large-scale 3D point cloud data containing millions of cell spots in real-time with a smooth frame rate of 60 frames per second or higher.

The SpatialVista ecosystem developed by the research team is provided in three interfaces tailored to the researcher's analysis environment. The first is a Python package that is directly implanted into the Jupyter environment, which is mainly used by data scientists. The second is a web service that can be launched immediately in a web browser without any separate installation process, and it can also be downloaded as a standalone desktop program. As a result, data analysts can easily confirm 3D visualization results instantly during coding and immediately reflect feedback into the analysis.

In experiments using actual mouse brain tissue and human tumor biopsy samples, SpatialVista demonstrated excellent data manipulation capabilities. Analysts can freely rotate or zoom in on the 3D cell map using mouse manipulation. It also supports a slicing function that cuts through specific tissue areas like a knife to view the internal structure. It is also possible to dynamically adjust the transparency and size of cells according to the threshold of the expression level of a specific gene, which is advantageous for intuitively distinguishing key cell groups within complex tissues.

Significance and Prospects

SpatialVista is significant in that it solves the biggest obstacles in high-resolution 3D ST research, which were computational delays and analysis disconnections. It has opened the way for clinicians and biologists who are not familiar with coding to easily explore the stereoscopic tissue microenvironment using an intuitive graphical interface. In particular, it is expected to activate research that visually confirms the 3D distribution of blood vessels in cancer tissues and the tumor infiltration pathway of immune cells. The visual elucidation of the stereoscopic interactions between cells can be a key to improving the efficiency of targeted drug development.

However, the fact that 3D ST technology itself is still in the early stages of standardization is a challenge to be addressed. The key is how to correct the slight errors that occur when aligning data produced on different devices or platforms. The research team plans to embed artificial intelligence algorithms into the system in the future to improve the 3D alignment accuracy between data from different devices. It remains to be seen whether this tool can become a standard platform for popularizing high-dimensional biological data analysis.