Accurate, high-throughput spatial profiling of whole slide samples with the Paletrra™ multiplexed image analysis pipeline
The tumor microenvironment is increasingly recognized as an important predictor of patient response to immunotherapy, but reading it across a whole slide at single-cell resolution is only as good as the image analysis behind it; this AACR 2025 poster from NeoGenomics benchmarks its Paletrra pipeline against clinically validated immunohistochemistry on sequential slides, reporting concordance of 0.85 to 0.97 across the six markers evaluated.
Presented by NeoGenomics.
Read or download to see:
- Concordance with clinically validated IHC ran 0.85 to 0.97 across the six markers benchmarked. Sequential slides were stained by immunofluorescence and by IHC for CD3, CD8, CD163, Ki67, PanCK and PD-L1; density and percent-positive figures are reported for all six, and H-score for Ki67 and PD-L1 only. The PD-L1 tumor H-score comparison is plotted as a scatter labelled r = 0.97.
- The 16-marker panel was built over eight staining rounds of two dyes each, imaged on the RareCyte CyteFinder II HT. Each round applied a pair of conjugated fluorescent antibodies on Cy3 and Cy5 before the dye was erased for the next pair, and cells were detected from the round 1 DAPI stain and tracked across rounds. Paletrra itself is a NeoGenomics multiplexed immunofluorescence service for up to 60 proteins on a single FFPE section, over an imageable area of nearly 10 square centimeters.
- Neighborhood clustering across 19 NSCLC samples grouped the neighborhoods into nine labelled types. Marker prevalence was tallied within each ten-cell neighborhood and clustered by K-means, then stacked per sample and ordered by PD-L1 positivity; the poster reads three of those samples as T-cell active, T-cell excluded and T-cell desert. Each biomarker has its own dedicated binary classifier, the models are trained on millions of annotated cells in a human-in-the-loop workflow guided by an in-house clinical pathologist, and single cells are phenotyped with 100% scientist QC.
