Single-cell phenotypic analysis and multiplet detection through incorporation of microscopy data into cellenONE-based single-cell proteomic data analysis

The reliability of single-cell proteomics (SCP) is intrinsically linked to the fidelity of cell isolation; however, the identification of co-isolated cells (doublets or multiplets) remains a persistent challenge for single-cell (sc-) omics. While single-cell transcriptomics has established probabilistic frameworks for doublet detection, these methods are ill-suited for the sparser throughput of SCP datasets. This work presents scpImaging , a novel computational pipeline that repurposes the latent microscopy data routinely generated during cellenONE-based sample preparation to provide deterministic quality control (QC) and high-content phenotypic profiling. We demonstrate that standard proteomic quality metrics for SCP (e.g., peptide count, signal intensity) paradoxically favour retaining doublets. In contrast, scpImaging applies machine-learning-based cell segmentation (Cellpose-SAM) to identify and exclude these artefacts with high precision. Furthermore, the framework integrates morphological metrics generated using the open-source CellProfiler, with proteomic abundance data, enabling joint analysis that links cell shape and texture to molecular cell state. Provided as an open-source R package, scpImaging offers a scalable, automated solution for enhancing data integrity and adding a phenotypic dimension to SCP experiments without increasing experimental cost or complexity.