MGCL-ST: multi-view graph contrastive learning for spatial transcriptomics imputation

High-resolution, continuous spatial gene expression profiling is critical for dissecting complex tissue microenvironments, yet current spatial transcriptomics (ST) platforms suffer from spatial gaps and insufficient resolution. Here, we present MGCL-ST, a multi-view graph contrastive learning method for super-resolution ST imputation. By jointly modeling local and global spatial graphs with histological features derived from a pathology foundation model, MGCL-ST accurately imputes unmeasured gene expression across the tissue space. Evaluated across three diverse platforms, MGCL-ST outperforms state-of-the-art methods in imputation accuracy and spatial clustering, significantly enhancing biological interpretability and enabling the precise mapping of tumor microenvironments. MGCL-ST is available at https://github.com/guxin2002/MGCL-ST .