The Xenium Analyzer run is complete and it is time to start the in-depth analysis of your results. This article will provide guidance to aid you in the next steps of your journey as you learn how to analyze your in situ data.
This guide will cover how to to perform certain tasks such as:
- Data quality control (QC)
- Data visualization with Xenium Explorer
- Alternative cell segmentation tools
- Analysis with community-developed tools
Here, we provide some tools to help you explore and prepare your data for additional analyses. This tutorial discusses Xenium Analyzer, Xenium Explorer, and community-developed tools.
Examining the quality of your data is always a good place to start in your analysis journey. We have documentation on checking data quality on our support site here. The Phred-scaled quality score (Q-Score) is the first metric you should examine. This metric is used to assess the quality of decoded transcripts. It indicates the probability that a given transcript is correctly identified by the decoding algorithm in the Xenium Onboard Analysis pipeline. The Q-Score is influenced by technical factors such as signal brightness, spot localization accuracy, and signal purity. There are control probes that are built into the system to ensure that the reported Q-Scores are accurately calibrated.
- Unassigned codewords: unused codeword - there is no probe in this particular gene panel that will generate the codeword
- Negative Control Codeword: features are codewords in the codebook that do not have any probes that should match that code, so they can be used to assess the specificity of the decoding algorithm
- Negative Control Probes: features are probes that exist in the panels, but target non-biological sequences, which can be used to assess the specificity of the assay
The cell-feature matrix and Xenium Analyzer’s secondary analyses only include transcripts with a Q-Score ≥ 20.
Xenium Analyzer generates files in open formats that can be opened by Xenium Explorer. It is a desktop application that allows interactive in situ data visualization, including the morphology image, RNA transcripts, and segmented cells. It enables interactive and dynamic exploration of tissue macrostructure and subcellular transcript localization. You can pinpoint the locations of specific transcripts, examine cell segmentation, and compare gene expression profiles and cell type proportions across tissue regions. Download Xenium Explorer here and find tutorials here.
A critical step in Xenium data analysis is the identification of cell nuclei and cell borders. Cell segmentation is part of the standard Xenium Onboard Analysis pipeline where cell segmentation is inferred based on the images collected during sample processing. For reference, Xenium Analyzer’s cell segmentation algorithm is described in this knowledge base article. You might be interested in exploring different cell segmentation tools as alternative approaches.
Cellpose and Baysor are two commonly used community-developed cell segmentation tools. Cellpose uses a generalist algorithm for 3D nucleus segmentation, and we have an Analysis Guide tutorial on how to use this tool on Xenium Analyzer results here.
Baysor optimizes cell segmentation based on transcriptional composition. It also has the capability to integrate stains such as DAPI. We have a tutorial here that demonstrates how to use this tool on Xenium Analyzer results.
You may wish to perform an analysis that is beyond what is currently supported by the 10x Genomics Xenium Analyzer platform. Luckily there is a robust software developer environment and a rapidly evolving set of community-developed tools that support the analysis of 10x Genomics Xenium Analyzer data. These tools require some programming knowledge, such as R and Python. Here is a short list of popular Xenium-compatible community-developed tools and some of the analyses they support:
Community-developed tools written in R
- Seurat: QC, cropping images and cell boundaries, and unsupervised clustering
- Giotto: add statistics, normalize expression, calculate high variable features, dimension reduction, clustering, and subcellular visualization
- Voyager: QC, spatial autocorrelation of QC metrics, Moran’s I (correlation coefficient, measures how one object is similar to others surrounding it), dimension reduction, differential expression, and local spatial statistics of marker genes
Community-developed tools written in Python
- stLearn (two tutorials, CCI, PSTS): normalization, gridding, permutation test for high co-expression, cell-cell interactions, clustering, and pseudo-time-space spatial trajectory analysis,
- Squidpy: QC, dimension reduction, spatial statistics, neighbors enrichment analysis, and compute Moran’s I score
In addition, there are other community-developed tools for spatial transcriptome analysis that are not Xenium-centric:
- RCTD (R): cell type identification, cell type-specific differential expression analysis, sample aggregation, and batch correction analysis
- MERINGUE (R): encode spatial relationship, identify spatial expression heterogeneity, identify gene expression patterns indicative of cell-cell communication with the capability to accommodate 2D and 3D spatial data
- SCIMAP (Python): build neighborhood graft, identify recurrent cellular neighborhood, cell type calling, adding ROI, and Napari visualization
Image viewer community-developed tools
The first preprint using the Xenium Analyzer is an excellent starting point if you are interested in seeing one example of a Xenium Analyzer analysis workflow:
The list of tools above is not a comprehensive summary. The landscape of Xenium data analysis is rapidly evolving. New and exciting in situ spatial transcriptome analysis tools, articles, and resources continue to be announced as the field grows. We hope this article has inspired you on your Xenium analysis journey by identifying some sites to see along the way.