Spatially resolved transcriptomics (SRT) is a groundbreaking technique in biology that allows scientists to map and understand the complex structures and functions of tissues at a detailed level. As SRT studies become more common and large-scale, the need for efficient and accurate methods to analyze this data grows. Enter IRIS (Integrative and Reference-Informed Tissue Segmentation), a new computational method designed to make sense of SRT data by detecting spatial domains within tissues.
What is IRIS?
IRIS is a tool developed by researchers at Brown University and the University of Michigan that helps researchers analyze SRT data by identifying and characterizing different regions within tissues. It does this by using reference data from single-cell RNA sequencing, which provides detailed information about the types of cells present and their gene expression patterns. By leveraging this reference data, IRIS can more accurately identify biologically meaningful spatial domains within the tissue.
How Does IRIS Work?
IRIS integrates data from multiple SRT slices, taking into account the relationships within each slice and across different slices. This integration allows for a more comprehensive analysis of the tissue. Essentially, IRIS looks at the data from various angles and combines it to create a clearer picture of the tissue’s structure and function.
Schematic overview of IRIS
IRIS is an accurate and efficient integrative reference-informed segmentation method for detecting spatial domains on multiple tissue slices across a range of SRT technologies. IRIS requires as input SRT data measured on multiple tissue slices with spatial localization information, along with scRNA-seq reference data measured on the same tissue with cell-type-specific gene expression information. With these two data inputs, IRIS performs integrative and reference-informed domain detection by integrating scRNA-seq data into the SRT data and further integrating SRT data across multiple tissue slices. Such integrative analysis of IRIS is carried out through a joint modeling framework, with an efficient iterative optimization algorithm for ensuring scalable computation. One unique feature of IRIS is its ability to leverage cell-type-specific expression profiles from the reference scRNA-seq data to facilitate the mapping of spatial domains in SRT studies. Another unique feature of IRIS is its ability to incorporate the spatial transcriptomic profiles from neighboring spatial locations on each single tissue slice as well as that across multiple tissue slices, while properly accounting for both within- and between-slice correlations, to achieve accurate and consistent spatial domain detection across slices. As a result, IRIS is highly accurate, robust and scalable.
The Advantages of IRIS
The creators of IRIS tested it on six different SRT datasets that included a variety of technologies, tissues, species, and resolutions. The results were impressive:
Accuracy: IRIS showed substantial improvements in accuracy, ranging from 39% to over 1,000% compared to other methods.
Speed: IRIS was significantly faster, with speed improvements ranging from 4.6 times to 666 times faster than existing methods.
Applicability: Unlike other methods, IRIS could handle large datasets, including those from advanced technologies like Stereo-seq and 10x Xenium.
Real-World Applications
IRIS has proven its worth in several practical applications:
Brain Structures: It revealed intricate details of brain structures, providing deeper insights into how different parts of the brain are organized and function.
Tumor Microenvironment: In cancer research, IRIS uncovered the heterogeneity within tumor microenvironments, which is crucial for understanding how tumors grow and respond to treatments.
Diabetes-Affected Testis: In studies of diabetes, IRIS detected structural changes in the testis, helping researchers understand how diabetes affects this tissue.
Conclusion
IRIS represents a significant advancement in the analysis of spatially resolved transcriptomics data. By combining accuracy, speed, and the ability to handle large datasets, IRIS opens up new possibilities for mapping and understanding complex tissue structures and functions. Whether studying brain development, cancer progression, or the effects of diseases like diabetes, IRIS provides researchers with a powerful tool to unlock the secrets hidden within our tissues.
Availability – The IRIS software package and source code have been deposited at https://xiangzhou.github.io/software/ and https://github.com/YingMa0107/IRIS.
Ma Y, Zhou X. (2024) Accurate and efficient integrative reference-informed spatial domain detection for spatial transcriptomics. Nat Methods 21(7):1231-1244. [article]