MSGPCA: Multi-Slice Graph PCA for replicate-aware Spatial Omics analysis
As spatial transcriptomics (ST) and spatial proteomics (SP) technologies mature, experimental designs are increasingly moving beyond single-slice analyses toward multi-slice studies involving one or more donors and experimental conditions. Although these designs enable the identification of reproducible spatial signals, they also introduce substantial biological heterogeneity, particularly when…
Spatial transcriptomics (ST) and spatial proteomics (SP) technologies have advanced to the point where experiments now often involve multiple slices and donors, examining various conditions. While these experiments help identify consistent spatial signals, they also introduce considerable biological variability, especially when dealing with non-sequential slices or areas with distinct anatomical structures.
If this variability is not properly accounted for, it can obscure slice-specific tissue architecture, obscure shared molecular patterns, and hinder the discovery of meaningful latent structures. Dimension reduction is crucial for condensing high-dimensional molecular data into a more manageable lower-dimensional space; however, most current multi-slice methods use a globally uniform representation that doesn't adequately handle slice-level variability.
To tackle this issue, we introduce Multi-Slice Graph Principal Component Analysis (MSGPCA), which breaks down molecular variation into two components: shared spatial factors that remain constant across slices, and slice-specific factors that capture local tissue microarchitecture. These MSGPCA-derived representations maintain spatial tissue structure, reduce noise in molecular profiles, and uncover biologically meaningful metafeatures related to shared and slice-specific biology.
In a mass spectrometry imaging dataset featuring non-sequential slices of ductal carcinoma in situ (DCIS) and invasive breast cancer (IBC), the shared factors identified broad biological differences between the two tissue regions, while the slice-specific factors highlighted intratumoral spatial variation within the IBC microenvironment.
When applied to human dorsolateral prefrontal cortex ST data, MSGPCA successfully captured the layered cortical architecture across adjacent slices, closely matching expert pathologist annotations. In summary, these results demonstrate that MSGPCA effectively isolates shared tissue architecture while preserving local microenvironmental variation in intricate multi-slice spatial omics datasets.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
