UniFORM: Towards Universal ImmunoFluorescence Normalization for Multiplex Tissue Imaging

Dataset Title: PRAD‑CyCIF: Prostate Cancer Single‑Cell Mean Fluorescence Intensity Dataset from Cyclic Immunofluorescence (CyCIF) — AnnData & Pickle Versions Description: This publicly released PRAD‑CyCIF dataset is a curated, single‑cell feature representation of prostate cancer tissues imaged on the Cyclic Immunofluorescence (CyCIF) platform. It underpins the benchmarking experiments in the manuscript “UniFORM: Towards Universal ImmunoFluorescence Normalization for Multiplex Tissue Imaging.” Two parallel data formats are provided to accommodate common workflows: AnnData Version (.h5ad) Expression matrix (.X):– Shape: 1,760,530 cells × 20 markers– Each row = one cell’s mean intensity across 20 markers Cell metadata (.obs): cell_id: unique cell identifier sample_id: patient sample (PRAD‑01 … PRAD‑20) scene_id: imaging scene label (e.g., scene000) batch_id: staining/imaging batch (e.g., batch1 … batch7) x, y: cell coordinates within each scene Marker metadata (.var): marker_name: the 20 marker panel (e.g., DAPI_R1, EPCAM, CD45, Ki67, etc.) Unstructured info (.uns['image_dimensions']):– Pixel height/width of each scene Pickle Version (.pkl per sample) Structure: a directory of 20 files, one per patient sample named PRAD‑XX_mean_intensity.pkl Sample order: the pickle file is saved in exactly this order: PRAD_sample_names = ['PRAD-01', 'PRAD-02', 'PRAD-03', 'PRAD-04', 'PRAD-05', 'PRAD-06', 'PRAD-07', 'PRAD-08', 'PRAD-09', 'PRAD-10', 'PRAD-11', 'PRAD-12', 'PRAD-13', 'PRAD-14', 'PRAD-15', 'PRAD-16', 'PRAD-17', 'PRAD-18', 'PRAD-19', 'PRAD-20'] Marker order: the channels are stacked in exact this order: PRAD_markers = ['DAPI_R1', 'EPCAM', 'CD56', 'CD45', 'aSMA', 'ChromA', 'CK14', 'Ki67', 'GZMB', 'ECAD', 'PD1', 'CD31', 'CD45RA', 'HLADRB1', 'CD3', 'p53', 'FOXA1', 'CDX2', 'CD20', 'NOTCH1'] Contents (per .pkl): { 'intensity_mean': np.ndarray of shape (20 markers, N_cells),} Intended Use: This dataset is intended for researchers to replicate results shown in our manuscript and further benchmarking and method development. Acknowledgments: Please cite the dataset and related publications appropriately when using this data in your research.