Evaluation of Cross-slice Imputation for Spatial Proteomics

Learning-based protein imputation methods significantly outperform alignment-based methods for cross-slice protein imputation, while inter-patient imputation is far less accurate than intra-patient imputation across evaluated methods.

• Why it matters: Cross-slice protein imputation addresses the limited biomarker coverage of spatial single-cell protein profiling by learning proteins unique to adjacent tissue slices.
• What they did: The study evaluates multiple learning-based and alignment-based approaches for CSPI using known single-cell biomarker expressions from physically adjacent biopsy slices with overlapping, non-subsuming marker sets.
• The result: Statistical performance analysis under varying implementation conditions shows learning-based methods are better for CSPI, whereas alignment-based methods fit inter-slice interpolation or 3D reconstruction, and inter-patient transfer is less accurate than intra-patient.