While we applied this framework to nPOP (v2)–based SCP, it is not dependent on any steps taken after cell dispensing. As scpImaging operates entirely on the imaging data produced by the cellenONE, it can be applied to any sc-omics workflow or cell-sorting experiment performed on the instrument.

The scpImaging R package (github.com/emmottlab/sc...), interactive Shiny application and the custom iSEE panel (emmottlab.shinyapps.io/scpImaging_H...) for exploring images, segmentation masks, morphology and proteomic data are all openly available!

To make this easier to explore alongside SCP data, we’ve included an interactive component. A custom iSEE panel links each cell’s image, segmentation, morphology and proteomic profile, making it straightforward to inspect outliers and check doublet removal.

These can be aligned with the proteomic measurements for each cell, providing a way to examine how morphological variation relates to molecular state. This adds a genuinely useful phenotypic layer to SCP without changing the underlying experimental workflow.

The second part focuses on the fact that these images contain much more than just cell counts. Using CellProfiler, we extract a broad set of morphological features (size, shape, texture, circularity, etc.).

In the first part of our work, we show that these images can be used directly to resolve this. We trained a Cellpose-SAM model to segment cells from cellenONE micrographs and used those masks to distinguish singlets from doublets with accuracy comparable and even exceeding manual inspection.

One of the persistent issues in SCP is the difficulty of identifying doublets from standard QC metrics alone. Doublets can appear as “high-quality” cells and distort downstream analyses.