CellFuse enables multi-modal integration of single-cell and spatial proteomics data

Fig 2

Bone marrow (Fig 2B, D, E, F, H, Supplementary Fig 1A, 2,3)

1.        Fig 2/BM/Reference/ Fig2_BM_prepare_data.R: Prepare bone marrow for CellFuse

2.        Fig 2/BM/ BM_CellFuse_Integration.R: Run CellFuse

3.        Fig 2/BM/BM_Running_Benchmark_Methods.R:  Run benchmarking methods (Harmony, Seurat, FastMNN)

4.         Fig 2/BM/BM_scIB_Benchmarking.ipynb: evaluate performance of CellFuse and other benchmarking methods using scIB framework proposed by Luecken et al.

5.        Fig 2/BM/ BM_scIB_prepare_figures.R: Visualize results of scIB framework

6.        Fig 2/BM/Sequential_Feature_drop/Prepare_data.R: Prepare data for evaluating sequential feature drop

7.        Fig 2/BM/Sequential_Feature_drop/Run_methods.R: Run CellFuse, Harmony, Seurat and FastMNN for sequential feature drop

8.        Fig 2/BM/Sequential_Feature_drop/Evaluate_results.R: Evaluate results features drop and visualize data.

PBMC (Fig 2G,I, Supplementary Fig 1B and 4)

1.        Fig 2/PBMC/Reference/ Fig2_PBMC_prepare_data.R: Prepare PBMC data for CellFuse

2.        Fig 2/ PBMC / PBMC_CellFuse_Integration.R: Run CellFuse

3.        Fig 2/ PBMC /PBMC_Running_Benchmark_Methods.R:  Run benchmarking methods (Harmony, Seurat, FastMNN)

4.         Fig 2/ PBMC /PBMC_scIB_Benchmarking.ipynb: evaluate performace of CellFuse and other benchmarking methods using scIB framework proposed by Luecken et al., 2021

5.        Fig 2/ PBMC /PBMC_scIB_prepare_figures.R: Visualize results of scIB framework

6.        Fig 2/ PBMC/ RunTime_benchmark/Run_Benchmark.R: Prepare data, run benchmarking method and evaluate results.

Fig 3 and Supplementary Fig 5

1.        Fig 3/Reference/ Fig3_CyTOF_prepare_data.R: Prepare CyTOF and CITE-Seq data for CellFuse

2.        Fig 3/CellFuse_Integration_CyTOF.R: Run CellFuse to remove batch effect and integrate CyTOF data from day 7 post-infusion

3.        Fig 3/CellFuse_Integration_CITESeq.R: Run CellFuse to integrate CyTOF and CITE-Seq data

4.        Fig 3/CART_Data_visualisation.R: Visualize data

Fig 4 

HuBMAP CODEX data (Fig. 4A, B, C, D and Supplementary Fig 6)

1.        Fig 4/CODEX_colorectal/Reference/ CODEX_HuBMAP_prepare_data.R: Prepare CODEX data from annotated and unannotated donor

2.        Fig 4/ CODEX_colorectal/ CODEX_HuBMAP_CellFuse_Predict.R: Run CellFuse on cells from from annotated and unannotated donor

3.        Fig 4/ CODEX_colorectal/CODEX_HuBMAP_Data_visualisation.R: Visualize data and prepare figures.

4.        Fig 4/ CODEX_colorectal/ CODEX_HuBMAP_Benchmark.R: Benchmarking CellFuse against CELESTA, SVM and Seurat using cells from annotated donors and prepare figures.

a.        Astir is python package so run following python notebook: Fig 4/ CODEX_colorectal/ Benchmarking/Astir/Astrir.ipynb

5.        Fig 4/ CODEX_colorectal/CODEX_HuBMAP_Suppl_figure_heatmap.R: F1score calculation per celltype per Benchmarking methods and heatmap comparing celltypes from annotated and unannotated donors (Supplementary Fig 6)

IMC Breast cancer data (Fig. 4E,F, G and Supplementary Fig 7)

1.        Fig 4/ IMC_Breast_Cancer/ IMC_prepare_data.R: Prepare CODEX data from annotated and unannotated donor

2.        Fig 4/ IMC_Breast_Cancer/ IMC_CellFuse_Predict.R: Run CellFuse to predict cell types

3.        Fig 4/ IMC_Breast_Cancer/ IMC_dat_visualization.R: Visualize data and prepare figures.

Fig 5

1.        Fig5/ Reference/ Fig5_CyTOF_Data_prep.R: Prepare CyTOF data from healthy PBMC and healthy colon single cells

2.        Fig5/ MIBI_CellFuse_Predict.R: Run CellFuse to predicte cells from colon cancer patients

3.        Fig5/ MIBI_PostPrediction.R: Visualize data and prepare figures

4.        Fig5/ Predicted_Data/ mask_generation.ipynb: Post CellFuse prediction annotated cell types in segmented images. This will generate Fig5C and D