Seurat assay metadata extraction

Seurat Assay Metadata Extraction

License: restricted — no clear open-source license detected for the underlying tool; verify licensing before commercial use or redistribution. <!-- asb-license-banner -->

Summary

Extract feature-level metadata (m/z values, annotation columns, and derived attributes) from a Seurat object's assay slot to enable downstream filtering, refinement, and statistical analysis of spatial metabolomics data. This skill bridges the gap between Cardinal MSI objects and Seurat-based annotation workflows by preserving and accessing feature-level metadata that would otherwise be inaccessible through standard Seurat accessors.

When to use

Use this skill when you have a SpaMTP Seurat object with a 'Spatial' assay containing metabolomics features (m/z values) and their associated metadata columns (e.g., 'all_IsomerNames', adduct types, or lipid nomenclature), and you need to: (1) extract all annotations for a given m/z to apply lipid simplification or nomenclature refinement; (2) count annotation cardinality per feature before and after refinement to validate reduction; (3) filter features by annotation presence/quality; or (4) prepare metadata for statistical testing or pathway association. This is essential when the original Cardinal object's metadata was transferred during CardinalToSeurat conversion and must be accessed or manipulated in R data frames.

When NOT to use

• Input Seurat object lacks a 'Spatial' assay or the assay contains no feature metadata (e.g., only intensities, no annotations) — use direct Cardinal object access instead.
• You only need to access intensity values or spatial coordinates, not feature-level annotations — use standard Seurat accessors (GetAssayData, GetTissueCoordinates) instead.
• The original Cardinal object is still available and you need to preserve its complete provenance — work directly with Cardinal assay metadata rather than the Seurat conversion.

Inputs

• SpaMTP Seurat object with 'Spatial' assay (containing normalized m/z intensities)
• Seurat assay slot containing feature metadata ([redacted-email])
• Annotation column name(s) as strings (e.g., 'all_IsomerNames')

Outputs

• R data frame with m/z values as row names and metadata columns (e.g., isomer names, lipid classes, adduct types)
• Optionally: filtered or transformed metadata for downstream analysis (e.g., count of annotations per m/z, simplified lipid nomenclature)

How to apply

Access the feature metadata of a Seurat spatial assay using object@assays$[redacted-email], which returns an R data frame indexed by feature names (m/z values as row names). Each row corresponds to one m/z feature; each column represents a metadata attribute (e.g., 'all_IsomerNames' containing semicolon-delimited isomer strings, or other annotation columns). Use standard R data frame operations (dplyr::select, tidyr::separate_rows, base R subsetting) to extract, filter, or transform specific columns. For example, to count annotations per m/z, split the semicolon-delimited strings using tidyr::separate_rows and count unique metabolite entries per m/z; to prepare for RefineLipids, pass the extracted metadata frame with the target annotation column specified (annotation.column='all_IsomerNames'). Validate the extraction by checking row count (should equal total m/z features in the object) and confirming that NA values are preserved for non-lipid or unannotated metabolites. The extracted data frame can then be manipulated independently or re-assigned to the assay slot after refinement.

Related tools

• SpaMTP (Provides CardinalToSeurat conversion and RefineLipids function for downstream refinement of extracted metadata) — https://github.com/GenomicsMachineLearning/SpaMTP
• Seurat (Object model and assay slot architecture for storing and accessing feature metadata) — https://satijalab.org/seurat/
• dplyr (Data frame manipulation and filtering (select, filter, mutate operations on extracted metadata))
• tidyr (Reshaping and splitting of semicolon-delimited annotation strings (separate_rows))
• Cardinal (Original MSI object model from which metadata is transferred via CardinalToSeurat) — https://github.com/Vitek-Lab/Cardinal3-vignettes

Examples

# Extract metadata from Seurat spatial assay and prepare for RefineLipids
refined_lipid_annotations <- RefineLipids(spotted@assays$[redacted-email], annotation.column='all_IsomerNames', lipid_info='simple')

Evaluation signals

• Row count of extracted data frame equals total number of m/z features in the Seurat object (verify with nrow(metadata_df) == nrow(GetAssayData(object, slot='counts')))
• Row names of extracted data frame match the feature names of the Seurat object (verify with identical(rownames(metadata_df), rownames(GetAssayData(object))))
• Annotation columns contain expected content: semicolon-delimited isomer strings, NA values for unannotated m/z, or simplified lipid nomenclature after RefineLipids application
• After splitting semicolon-delimited annotations (tidyr::separate_rows), the maximum count of unique annotations per m/z matches the reported reduction metric (e.g., 161 unique annotations before refinement)
• No unexpected NA patterns or all-NA columns, indicating metadata was correctly transferred during CardinalToSeurat conversion

Limitations

• Metadata preservation depends entirely on successful transfer during CardinalToSeurat conversion; if the original Cardinal object lacked complete metadata, extraction will yield incomplete data.
• Semicolon-delimited annotation strings (e.g., 'all_IsomerNames') must be consistently formatted; inconsistent delimiters or embedded semicolons will cause splitting errors during downstream refinement.
• NA values in feature metadata are preserved but require explicit handling (e.g., na.omit or dplyr::filter(!is.na(column))) to avoid errors in count operations or refinement functions.
• The article notes that the CardinalToSeurat conversion module is incompletely documented ('Unable to produce a finding — the provided section text contains only a README header'); exact behavior regarding edge cases (e.g., non-lipid metabolites, missing spatial coordinates) is not fully specified.

Evidence

• [methods] Extract the Spatial assay metadata containing the 'all_IsomerNames' annotation column.: "Extract the Spatial assay metadata containing the 'all_IsomerNames' annotation column."
• [methods] Call RefineLipids() on the assay metadata with annotation.column='all_IsomerNames' and lipid_info='simple' to simplify lipid nomenclature into Lipid Maps categories and classes.: "Call RefineLipids() on the assay metadata with annotation.column='all_IsomerNames' and lipid_info='simple' to simplify lipid nomenclature"
• [methods] Preserve feature metadata (m/z values) in the feature meta.data slot.: "preserve feature metadata (m/z values) in the feature meta.data slot"
• [other] Runs lipid nomenclature simplification on annotations: refined_lipid_annotations <- RefineLipids(spotted@assays$[redacted-email]: "refined_lipid_annotations <- RefineLipids(spotted@assays$[redacted-email]"
• [methods] Count the number of annotations per m/z before refinement by splitting the semicolon-delimited isomer name strings.: "Count the number of annotations per m/z before refinement by splitting the semicolon-delimited isomer name strings."