A study published in Nature Biomedical Engineering introduces an automated, open-source pipeline designed to streamline the analysis of multiplex immunostaining in cancer tissue at single-cell resolution. The method, called MARQO (multiplex-imaging analysis, registration, quantification and overlaying), aims to address the challenges of analyzing increasingly complex multiplex immunohistochemistry (IHC) and immunofluorescence (mIF) datasets.
The MARQO pipeline integrates elastic image registration, iterative nuclear segmentation, unsupervised clustering, and user-guided cell classification within a single system. It is designed to analyze whole-slide images from both singleplex IHC and multiplexed assays. The workflow begins with quality control and tissue masking, followed by automated segmentation and quantification. MARQO uses parallel computing to divide tissue into tiles, enabling large-scale datasets to be processed efficiently.
Validation studies compared automated results with manual annotations by pathologists. In hepatocellular carcinoma samples, automated segmentation achieved close alignment with manual counts, with an average Dice score of 83 percent. The system’s cell classification also performed at a level comparable to manual annotations across a range of immune and tumor markers, including CD3, FOXP3, CD68, and PanCK.
The MARQO pipeline was tested on tissue microarrays, biopsies, and resections from several tumor types, including lung, colorectal, and melanoma samples. Results correlated strongly with manual pathologist quantification, suggesting it could be applied across different tissue contexts.
To explore its clinical utility, the pipeline was applied to samples from a trial of immunotherapy in hepatocellular carcinoma. The analysis showed higher levels of CD8-positive T cells in patients who responded to treatment, consistent with known associations between immune infiltration and checkpoint inhibitor response. It also mapped spatial relationships between immune and tumor cells, providing additional context for interpretation.
The study emphasizes that the MARQO tool is suitable for research and preclinical work but is not yet ready for routine diagnostic use. Limitations include the need for high computational capacity and continued pathologist oversight for classification. The authors note that refinement and broader validation will be required before it could support routine clinical workflows.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
