Researchers develop new ways to reduce noise in spatial transcriptome data

Spatial transcriptomics is a high-throughput sequencing technique that has become increasingly popular in recent years. Spatial transcriptomes can provide sequencing location information and even corresponding pathological image data while providing gene expression profiling data similar to single-cell sequencing data. However, there is a lot of noise in gene expression profiling data for spatial transcriptome techniques, particularly the latest high-resolution spatial transcriptome techniques. This noise comes from the diluted low-throughput sequencing depth at each sequencing site and the additional experimental steps performed to preserve the sequencing position. These noises include, but are not limited to, drop-out phenomena in single-cell sequencing data. This noise creates a huge obstacle for researchers to extract valuable information from valuable spatial transcriptome data.

On August 4, 2022, Dr. Tao Wang’s team from the Center for Quantitative Biology at the University of Texas Southwestern Medical Center and Professor Wang Li from the University of Texas at Arlington published an article in the journal Nature Methods titled “Sprod for De-noising Spatial Transcriptomics Data Based on Position and Image Information.” article. The team reported on their invention of the Sprod method, which uses spatial location and pathological image information specific to spatial transcriptome data to correct for noise in spatial transcriptome gene expression profiling data.

During Sprod’s noise reduction process, each site of sequencing borrows gene expression information from nearby sites. Sites that are more similar to each other on pathological images (spatial transcriptome techniques with images, such as Visium) or on the overall transcriptomics profile (for spatial transcriptome techniques without images, such as Slide-Seq) and are sequenced in adjacent locations will borrow more information from each other. Based on this principle, Sprod constructed a latent Graph Model (Fig 1) that placed sequencing sites into this hidden map by analyzing the spatial distances and gene expression profiling characteristics of different sequencing sites. The expression spectrum information of the spatial transcriptome flows according to this hidden plot, thereby achieving noise reduction of the expression spectrum data. Sprod can be applied to a variety of spatial transcriptome technologies such as Visium, Slide-Seq, HDST, Seq-Scope, etc. The higher the resolution of the type of technology, the louder the noise, and the greater the role of the Sprod.

Fig 1: How the mathematical model inside the Sprod software works

Wang Tao’s group then verified the reliability of Sprod noise reduction on different spatial transcriptome technology datasets. For example, in Fig. In 2, Wang Tao’s group demonstrated the effects of using Sprod before and after noise reduction on a Visium dataset for ovarian cancer. This dataset provides immunofluorescence staining (IF) images corresponding to the spatial transcriptome, which includes CD45. Such as Fig. 2 As shown on the left, the RNA expression data of the gene PTPRC (PTPRC’s translation product cd45) in the CD45 IF and Vision data are very poorly matched, but after using Sprod (for independent verification purposes, the CD45 channel of the IF image was eliminated during the calculation of Sprod), the gene expression of PTPRC and the staining intensity of CD45 coincided well. In addition, Wang Tao’s research group also compared Sprod with the drop-out correction methods commonly used in scRNA-seq data analysis (scImpute and SAVER), and showed that Sprod was better than scImpute and SAVER in noise reduction.

Fig 2: The degree of coincidence between gene expression of PTPRC and immunofluorescence staining of CD45. Left: Raw Visium data; Right: Data after Sprod noise reduction

Wang Tao’s research group further applied Sprod to a series of Visium, Slide-Seq, Seq-Scope and other spatial transcriptome datasets, verifying that Sprod can effectively reduce noise for various data. The data after noise reduction have biologically more biologically relevant results in downstream analysis such as differential expression analysis, pathway enrichment, and cell-to-cell communications. The drop-out correction method for single-cell sequencing data simply uses the expression spectrum itself to correct problems in the expression spectrum data. This creates a phenomenon similar to overfitting or oversmoothing, which has received some criticism in academic circles. In contrast, Sprod leverages information unique to sequencing locations and case images in spatial transcriptome data. With this independent information, Sprod can perform noise reduction operations more precisely.

Taken together, spatial transcriptome technology provides a powerful tool for biomedical research. The analysis of spatial transcriptome data has become increasingly challenging as technology evolves. Wang Tao’s research group believes that rigorous data preprocessing is the key to correctly analyzing and understanding spatial transcriptome data, and Sprod noise reduction is an important and powerful part of preprocessing.

The total of the papers is Dr. Wang Yunguan and Dr. Song Bing. Other lead authors of the paper include Professor Xie Yang, Professor Xiao Guanghua, and Assistant Professor Wang Shidan of Southwest Medical Center. The University of Texas Center for Quantitative Biology Has Several Postdoctoral Recruitment Positions (,, Welcome bioinformatics talents from all majors to join us. (Source: Science Network)

Related paper information:

Special statement: This article is reproduced only for the purpose of disseminating information, and does not mean to represent the views of this website or confirm the authenticity of its contents; If other media, websites or individuals reprint from this website, they must retain the “source” indicated on this website and bear their own legal responsibilities such as copyright; If the author does not wish to be reprinted or contact us for reprint fees, please contact us.

Source link

Related Articles

Leave a Reply

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

Back to top button