Background In immuno-oncology, accurate and actionable assessment of cellular mechanisms and interactions is paramount. The clinically validated multiplex immunofluorescence (mIF) assays at CellCarta accompanied by digital assays powered by RevealAI’s interactive platform offers a solution for clinical trials, preclinical research, and decision support in the immuno-oncology domain. The unique aspects of this solution lie in CellCarta’s rigorous validation process as well as the integration of AI-powered algorithms with an intuitive user interface, transforming the tumor microenvironment (TME) data in mIF images into actionable spatial insights.

Methods Various tumor samples were processed using Ultivue Immuno8 FixVUE™ and an Akoya MOTiF™ panel (CD8, CD68, FoxP3, PD-1, PD-L1, PanCK). The images were analyzed using RevealAI mIF platform to generate contextual and spatial biomarkers. Following the InSituPlex^® staining method for UltiVue panels and Akoya Opal staining protocol, FFPE samples were stained and scanned with the PhenoImager HT scanner (Akoya Biosciences).

Digital assays were developed to characterize and quantify T-cell subsets, immunosuppressive cells, and cellular interactions such as the PD-L1/PD-1 checkpoint axis in the tumor microenvironment. More specifically, MaskRCNN with post-processing of fluorescence images was used to improve the quality of nuclear segmentation (i.e., compared to StarDist and U-Net), signal preprocessing was employed to reduce background noise and autofluorescence by normalizing fluorescence intensity in different channels, and phenotype classification was performed using dynamic thresholding or neural network classifiers when necessary. The mIF assays and the accompanying results were validated by establishing concordance with pathologist readouts.

Results Quantitative and spatial results with validation procedures are presented. The spatial results include examples of spatial metrics that capture interactions in cell neighborhoods (i.e., FoxP3 and PD-L1 interactions with CD8 cells within a 30 um radius). Examples of quantifiable metrics to evaluate dispersion and arrangement of immune effector cells (CD3, CD4, and CD8) in the TME are provided, which allows automated sample stratification into Desert, Excluded, or Inflamed categories.

Conclusions The CellCarta’s standardized operating procedures have been successfully applied to validate the Ultivue and Akoya panels for sensitivity, specificity, and precision across various human samples. The RevealAI user-interactive platform was leveraged to generate quantitative and spatial metrics from the mIF images. This advanced platform enables the conversion of intricate multiplex imaging data into well-defined, spatial phenomic endpoints, supporting translational and clinical efforts and streamlining the biomarker discovery process.