MATHEMATICAL SCIENCES

Which spatial transcriptome analysis algorithm is stronger? Scientists systematically evaluate a variety of spatial transcriptome analysis algorithms


The qu kun research group of professor Qu Kun of the Department of Life Sciences and Medicine of the University of Science and Technology of China designed a complete set of analysis processes to systematically evaluate the performance of 16 spatial transcriptome and single-cell transcriptome data integration algorithms in predicting the spatial distribution of genes or cell types. The results of the research were published online on May 16 in Nature Methods.

Integrate analytics processes .png

Integrated analysis process Courtesy of The University of Science and Technology of China

“The spatial position of a cell within tissues and organs is closely related to its physiological function or the process of disease generation.” Guo Chuang, co-first author of the paper and a special associate researcher of the School of Life Sciences and Medicine of the University of Science and Technology of China, introduced it.

In recent years, researchers have developed a variety of spatial transcriptome techniques to detect the expression of the whole transcriptome within cells while retaining the precise spatial localization of cells, so as to study the cell subsets and their molecular mechanisms that play a key role in the development or development of disease.

However, there are two shortcomings in the current spatial transcriptomics technology: one is that the spatial transcriptome technology based on sequencing cannot achieve true single-cell resolution; the other is that the gene flux detected by the imaging spatial transcriptome technology is limited.

To push the limits of technology, bioinformatics scientists have designed multiple algorithms to integrate spatial transcriptome and single-cell transcriptome data to predict the spatial distribution of cell types and/or complete transcriptome information about individual cells. These algorithms have greatly deepened the understanding of spatial transcriptomics data and related biological and pathological processes.

However, there are significant differences in the working principle and scope of application of these integrated algorithms, and researchers have difficulty choosing the best algorithm.

Qu Kun’s research group has long been committed to the development of biological big data analysis algorithms and software. In this study, the research group collected 45 pairs of spatial transcriptome and single-cell transcriptome datasets from the same tissue source, 32 simulated datasets, and designed a variety of indicators, systematically evaluating the performance of 16 integrated algorithms from multiple dimensions such as accuracy, robustness, and time-consuming computing resources.

The results showed that cell2location, SpatialDWLS, and RCTD algorithms could predict the spatial distribution of cell types more accurately, while tangram, gimVI, and SpaGE algorithms were the best algorithms for predicting the spatial distribution of gene expression. Tangram, Seurat, and LIGER are relatively computationally efficient and suitable for working with large-scale data sets.

The research work summarizes the properties, performance and applicability of each algorithm, summarizes the advantages of efficient algorithms, and provides a reference for researchers to further improve the performance of algorithms. An analytical flow that integrates spatial transcriptome and single-cell transcriptome data is also provided on github, helping researchers choose the best analysis tool for processing their own data.

Ben Raphael, a professor at Princeton University, an expert in the field, commented, “It is not easy to conduct such a rigorous benchmark study of various spatial transcriptome data and methods, and this study meets important needs in the field. (Source: China Science Daily Wang Min)

Related paper information:https://doi.org/10.1038/s41592-022-01480-9



Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button