Antibody-guided chromogenic adaptation for pathology-semantic consistent hematoxylin and eosin to immunohistochemistry translation

Traditional immunohistochemical (IHC) staining provides essential biomarker information for clinical diagnosis but remains limited by high reagent cost, antibody-dependent optimization and labor-intensive workflows. In contrast, hematoxylin and eosin (H&E) staining is routine and inexpensive yet lacks molecular specificity. Recent advances in virtual staining offer the opportunity to computationally infer IHC appearances directly from H&E slides, thereby bridging the morphological-molecular gap while improving workflow efficiency. This study proposes a controllable multi-antibody virtual staining framework, termed antibody-guided staining adaptation (ASA), which synthesizes four major biomarkers (ER, HER2, Ki67 and PR) from directly H&E images by jointly modeling antibody semantics and chromogenic physics. Within ASA, antibody-conditioned response adjustment (ACRA) provides antibody-specific global modulation through sparsity-regularized and orthogonality-preserved feature transformations; spatial chromogenic adaptation (SCA) performs spatially coherent chromogenic adjustment by leveraging optical density (OD) derived cues; and patch-replacement dynamic switching (PRDS) introduces curriculum-based real-pseudo OD patch replacement to stabilize training under H&E-IHC misalignment. Extensive experiments show ASA generally improves structural and chromogenic quality across most biomarkers, with some metric-specific variations, pointing to its future clinical potential. Our code will be available at https://github.com/youzzz1203/ASA.HIGHLIGHTSDevelops ASA, a virtual staining framework using antibody and OD-driven modeling.Introduces ACRA, SCA and PRDS to enhance robustness under section misalignment.Achieves competitive IHC stain quality across multiple biomarkers.Enables efficient digital pathology workflows via virtual staining.Demonstrates potential as a clinical complement pending further validation.