spatial airr data

[grst]

Description of feature

Apparently this is a thing now (e.g. 10x spatial transcriptomics + mixcr, e.g. in this paper).

@giovp is there anything we need to do to leverage squidpy + scirpy together for this kind of data?
Can we account for this type of data in the design of the new data structure (#327)?

CC @FFinotello

[gtca]

One thing to add here would be that in case this is to be represented as a MuData object, we would like to provide an interface for this in muon.spatial relying on squidpy.

[grst]

https://twitter.com/soph_liu/status/1579913561215639554

[grst]

https://www.biorxiv.org/content/10.1101/2022.11.22.516865v1
https://www.biorxiv.org/content/10.1101/2023.04.01.535228v1

[grst]

https://www.science.org/doi/10.1126/science.adf8486

[grst]

From https://www.nature.com/articles/s41592-024-02243-4 (review on single-cell immune repertoire analysis)

Although there is considerable interest in spatial profiling of TCR/BCR, this field is very new. Parenthetically, downstream analysis options are limited. A common approach to spatial TCR-seq is by using MiXCR14 to perform TCR reconstruction from the spatial sequencing data generated from full-length cDNA libraries. Engblom et al.96 developed two protocols for spatial TCR/BCR-seq (Spatial VDJ) to capture paired-chain BCR and TCR sequences, as an extension of the 10x Genomics Visium spatial transcriptomics platform. The long-read version is compatible for capturing full-length BCR and TCR sequences, while the short-read version is only for TCR96. For the long-read Spatial VDJ data, relevant reads were prepared into fastq files and MiXCR was used for gene annotation. Short-read Spatial VDJ data were processed using pRESTO99 before analyzing with MiXCR. Spatial barcodes were appended to the MiXCR outputs using custom R scripts, and downstream clonotyping analysis was primarily achieved by leveraging the Immcantation19 workflow. Similarly, the recently described SPTCR-seq95 method includes a bespoke probe-based library preparation protocol that leverages the fact that several commercially available spatial transcriptomics technologies generate full-length cDNA libraries at some point in their respective protocols, allowing them to sensitively and specifically capture and reconstruct full-length TCR sequences using long-read sequencing. SPTCR-seq also includes a computational pipeline to extract the TCR-seq data, which includes annotation with IgBLAST against IMGT references and their SPATAimmune R package to import and visualize the SPTCR-seq data. Overall, while we note that capturing spatial TCR/BCR information can potentially provide additional resolution of spatially relevant immunological partners and immune response, there is a general lack of dedicated computational frameworks or tool kits that can carry out quality-control checks or perform integrated analysis of spatial information, RNA and TCR/BCR data. It also remains to be determined whether the appropriate data structures are available or suitable for analyzing these new data modalities.

[grst]

For supporting spatial AIRR data we need to distinguish between

• spot-based (visium-like) data and
• data with cellular or subcellular resolution.

For spot-based data, several steps in the workflow need to be adapted which is detailled in #583. For spatial data with single-cell resolution, the existing workflow should work as is, but we'd need to think about how to integrate well with squidpy and SpatialData.