Robust phenotyping of highly multiplexed tissue imaging data using pixel-level clustering (data)

MIBI-TOF data for lymph node dataset reported in Liu et al., Robust phenotyping of highly multiplexed tissue imaging data using pixel-level clustering

1. mibi_single_channel_tifs.zip: Single-channel MIBI-TOF images

Folders are labeled according to the field-of-view (FOV) number. Each folder contains single-channel TIFFs for each marker in the panel. Images are 1024x1024 pixels, 500 um. See paper for details.

2. segmentation.zip: Segmentation output of MIBI-TOF images

Cell segmentation was performed using Mesmer (Greenwald NF, Nature Biotechnology 2021). Output of Mesmer that delineates the single cells in each of the images is included.

3. source_data.zip: Source data files for figures

• pixel_ccs_allpreprocessing.csv: Cluster consistency score (CCS) for all pixels using all preprocessing steps, related to Fig. 2d-f, Supp. Fig. 4,5,9,10
• pixel_ccs_nopixelnorm.csv: CCS for all pixels where pixel normalization was left out, related to Fig. 2f, Supp. Fig. 6
• pixel_ccs_nochannelnorm.csv: CCS for all pixels where channel normalization was left out, related to Fig. 2f, Supp. Fig. 8
• pixel_ccs_passes1.csv: CCS for all pixels where 1 pass was used for SOM training, related to Supp. Fig. 10g
• pixel_ccs_passes100.csv: CCS for all pixels where 100 passes were used for SOM training, related to Supp. Fig. 10g
• pixel_ccs_sigma0.csv: CCS for all pixels where a Gaussian blur sigma of 0 was used for preprocessing, related to Supp. Fig. 5d
• pixel_ccs_sigma1.csv: CCS for all pixels where a Gaussian blur sigma of 1 was used for preprocessing, related to Supp. Fig. 5d
• pixel_ccs_sigma3.csv: CCS for all pixels where a Gaussian blur sigma of 3 was used for preprocessing, related to Supp. Fig. 5d
• pixel_ccs_sigma0_reps100.csv: CCS for all pixels where a Gaussian blur sigma of 0 was used for preprocessing and 100 replicates were used for CCS calculation, related to Supp. Fig. 5e
• pixel_ccs_sigma1_reps100.csv: CCS for all pixels where a Gaussian blur sigma of 1 was used for preprocessing and 100 replicates were used for CCS calculation, related to Supp. Fig. 5e
• pixel_ccs_sigma2_reps100.csv: CCS for all pixels where a Gaussian blur sigma of 2 was used for preprocessing and 100 replicates were used for CCS calculation, related to Supp. Fig. 5e
• pixel_ccs_sigma3_reps100.csv: CCS for all pixels where a Gaussian blur sigma of 3 was used for preprocessing and 100 replicates were used for CCS calculation, related to Supp. Fig. 5e
• pixel_ccs_nodes15.csv: CCS for all pixels where 15 nodes were used for SOM training, related to Supp. Fig. 9e
• pixel_ccs_threshold80.csv: CCS for all pixels where a threshold of 80% was used for CCS calculation, related to Supp. Fig. 4b
• pixel_ccs_threshold98.csv: CCS for all pixels where a threshold of 98% was used for CCS calculation, related to Supp. Fig. 4b
• pixel_info_comparison_table.csv: Number of pixels that were assigned to a cluster outside of cell segmentation masks, related to Fig. 3d
• single_cell_pixel_composition_table.csv: Pixel composition information for each single cell, related to Fig. 5, Supp. Fig 16
• single_cell_integrated_expression_table.csv: Integrated expression per cell, output by Mesmer, related to Fig. 5, Supp. Fig. 16
• cell_silhouette_scores.csv: Silhouette scores for comparing integrated expression and pixel composition, related to Fig. 5d
• cell_silhouette_scores_undefinedremoved.csv: Silhouette scores for comparing integrated expression and pixel composition where undefined cells were removed, related to Fig. 16g
• cell_silhouette_scores_preprocessed.csv: Silhouette scores for comparing integrated expression and pixel composition where pixels were preprocessed before integrating expression, related to Fig. 17d
• cell_silhouette_scores_ilastik_cellprofiler.csv: Silhouette scores for comparing integrated expression and pixel composition where segmentation masks were obtained using Ilastik/CellProfiler, related to Fig. 18b
• cell_ccs_pixel_composition.csv: CCS for all cells using pixel composition for clustering, related to Supp. Fig. 16e, 17c
• cell_ccs_integrated_expression.csv: CCS for all cells using integrated expression for clustering, related to Supp. Fig 16e-f
• cell_ccs_integrated_expression_preprocessed.csv: CCS for all cells using integrated expression for clustering where data was preprocessed before integrating, related to Supp. Fig 17c
• cytof_ccs.csv: CCS of the CyTOF dataset used as a benchmark, related to Supp. Fig. 4c,d
• scrnaseq_ccs.csv: CCS of the scRNA-seq dataset used as a benchmark, related to Supp. Fig. 4c,e
• pixel_phenotype_maps: TIFFs where pixel value corresponds to pixel cluster number as reported in the paper
• cell_phenotype_maps: TIFFs where pixel value corresponds to cell cluster number as reported in the paper
• runtime_analysis_pixel.csv: Runtime analysis of pixel clustering in Pixie, related to Supp. Fig. 22a
• runtime_analysis_cell.csv: Runtime analysis of cell clustering in Pixie, related to Supp. Fig. 22b
• runtime_clustering_algorithm.csv: Runtime analysis of different clustering algorithms, related to Supp. Fig. 22c