The role of computational pathology in predicting expression of selected immunohistochemical markers

Immunohistochemistry (IHC) remains a cornerstone of precision oncology, providing essential diagnostic, prognostic, and predictive molecular data. However, the traditional manual assessment of IHC slides faces persistent challenges regarding interobserver variability, reproducibility, and high diagnostic workloads. To address these limitations, artificial intelligence (AI) is increasingly integrated into computational pathology workflows. This paper outlines the current landscape of AI-assisted IHC, highlighting significant trends such as automated scoring, explainable AI, virtual staining, and multiplex analysis. We explore the transition from basic image analysis to advanced deep learning architectures capable of predicting specific IHC biomarker expression directly from standard hematoxylin and eosin morphology. Furthermore, we examine the commercial ecosystem of these tools and highlight the critical pre-analytical bottlenecks that hinder widespread clinical adoption. Finally, we present a practical case study demonstrating an automated, deep learning- based pipeline for quantifying CD34-positive myeloblasts in acute myeloid leukemia and myelodysplastic syndromes. Ultimately, while AI holds vast potential to optimize turnaround times and streamline laboratory triage, its successful clinical implementation will depend on seamless integration into existing digital pathology platforms and standardization of pre-analytical variables, moving the field from fragmented algorithms to reliable, standard-of-care diagnostic tools.