AnchorR: A QuPath and R interface for collaborative exploration of spatial transcriptomics and histology

AnchorR is a novel workflow designed to bridge the gap between single-cell spatial transcriptomics and histology, enabling accurate registration of tissue borders. In traditional analysis, tissue borders can vary due to sectioning artifacts, staining inconsistencies, and field-of-view differences, making conventional area-based registration challenging.

To address this, researchers created AnchorR, which employs an expert-guided approach for coarse-grained alignment between hematoxylin and eosin (H&E) images and CosMx Spatial Molecular Imaging data.

Bioinformaticians initiate the process by defining and color-coding cell types in Seurat, while pathologists subsequently identify corresponding internal landmarks using QuPath overlays. AnchorR then leverages these paired landmarks to estimate affine transformations, assess residual error, and facilitate visual quality control and anchor refinement.

In a pilot study involving six oral pre-cancerous tissue sections, the team identified 60 cross-modal landmarks. Remarkably, fitting each section independently reduced mean landmark error from 121.5 microns with a single whole-slide transformation to 14.6 microns. Furthermore, cross-validation demonstrated that incorporating more anchors enhanced robustness, with a nine-anchor fit achieving an error of approximately 20 microns, equivalent to roughly one cell diameter.

AnchorR serves as a valuable tool that complements automated computer-vision methods. By providing reliable tissue-level alignment in scenarios where global registration proves difficult due to border mismatches, it operationalizes an expert-in-the-loop approach. This approach makes feature-based multimodal registration accessible to researchers without requiring specialized computer-vision expertise or high-performance computing resources.