Mitigating bias in spatial transcriptomic pipelines via human feedback with CSDE

Biological discovery from large-scale assays requires extensive preprocessing, during which raw outputs (e.g., images, sequences) are processed into structured forms amenable to analysis. In spatial transcriptomics, differential expression analysis depends on cell segmentation, gene expression quantification, and cell-type annotation, all prone to systematic errors. We introduce corrected spatial differential expression (CSDE), building on prediction-powered inference to leverage a small set of expert-validated cells to correct for preprocessing bias. Across three case studies, CSDE yields more reliable differential expression estimates than the prevalent approach neglecting preprocessing.