The scripts for Wei et. al paper on cancer cell calling in colorectal cancer

Processed data is available from zenodo and should be unpacked into a directory structure like so (named ./data/):

.
├── cellbender
│   ├── p007n_cellbender_counts.h5
│   ├── p007t_cellbender_counts.h5
│   ├── p008n_cellbender_counts.h5
│   └── ...
├── cellranger
│   ├── p007n_raw_feature_bc_matrix.h5
│   ├── p007t_raw_feature_bc_matrix.h5
│   ├── p008n_raw_feature_bc_matrix.h5
│   └── ...
├── cellSNP
│   ├── downsampled
│   │   ├── p007t_05
│   │   │   ├── cellSNP.base.vcf
│   │   │   ├── cellSNP.samples.tsv
│   │   │   ├── cellSNP.tag.AD.mtx
│   │   │   ├── cellSNP.tag.DP.mtx
│   │   │   └── cellSNP.tag.OTH.mtx
│   │   ├── ...
│   ├── p007n
│   │   ├── cellSNP.base.vcf
│   │   ├── cellSNP.cells.vcf
│   │   ├── cellSNP.samples.tsv
│   │   ├── cellSNP.tag.AD.mtx
│   │   ├── cellSNP.tag.DP.mtx
│   │   └── cellSNP.tag.OTH.mtx
│   ├── ...
├── numbat
│   ├── p007_clone_post_2.tsv
│   ├── p008_clone_post_2.tsv
│   └── ...
└── WGS
    ├── p007_filtered.vcf.gz
    ├── p008_filtered.vcf.gz
    └── ...

• Ambient RNA background removal by CellBender
  • input: {sample}_raw_feature_bc_matrix.h5
  • output: {sample}_cellbender_counts.h5

• run 1_preprocessing_h5.ipynb
  • input: all cellbender_counts.h5 object
  • output:
    • CB_all_cells.h5 and CB_epi_cells.h5 that contains high-quality all/epithelial cells with calculated PCA, UMAP, diffusion map, and louvain embeddings, as well as coarse cell type annotation
    • anno/{sample}_{cell_type}.txt cell barcode list by sample and cell type

• run inferCNV/inferCNV.ipynb to execute inferCNV

  • input: CB_all_cells.h5
  • output: all_cell_CB_counts.rds, all_epi_cell_CB_counts.rds, all_cell_anno.txt, all_epi_cell_anno.txt, inferCNV_expression_data.rds and other inferCNV intermediate objects

• run inferCNV/inferCNV_result.ipynb to collect inferCNV result

  • collect inferCNV result from the previous run
  • output: infercnv_clone_scores.tsv

• run script/Snakefile for all the samples except p009
• run script//Numbat/combine_p009t1t2_and_run_numbat.ipynb since p009 have two normal samples and two tumour samples, they were run separately with modified scripts
• run script/Numbat/collect_numbat_result.ipynb to collect Numbat result of all the samples
  • output: numbat_all_output_clone_post_combined9.csv

• rub script/run_CCISM.sh to run CCISM on all the sample

• input
  • download the h5 object from Joanito et al.:
  • CB_epi_cells.h5
• run iCMS/preprocessing_Joanito.ipynb to filter the count matrix with the same criteria as our h5 objects
  • output adata_concat_with_joanito.h5
• run iCMS/run_scvi_snakemake.sh for scVI model training
• run iCMS/scvi_model_result_iCMS.ipynb to inspect scVI modeling result, train and inspect scANVI models
  • output `CB_epi_cells_iCMS.h5

• run 2_integrate_tools_result.ipynb
• input
  • CB_epi_cells_iCMS.h5 which contains iCMS result
  • CCISM result
  • Numbat: numbat_all_output_clone_post_combined9.csv
  • inferCNV: infercnv_clone_scores.tsv
• output: CB_epi_Numbat_CCISM_inferCNV_iCMS.h5

• run 3_cell_type_annotation_finer.ipynb
  • output adata_all_full_cell_type_annotation.h5

• run 4_consensus_calls.ipynb
  • output CB_epi_Numbat_CCISM_inferCNV_icms_Uhlitz_resolved_identity.h5

• run 5_cellrank.ipynb

• run 6a_DEG_epi.ipynb for expression and differential expression analysis of epithelial cells; 6b_DEG_imm_str.ipynb for immune and stromal cells

• run 7_CRC_pathway_signature.ipynb

• run 8_cellchat.ipynb to infer cell-cell interactions in the normal versus tumour samples

• run script/Figures.ipynb
  • input: data/data_consolidated.h5ad
  • output: figures/*
• Figure 2 contains data simulation result, run script/Fig2_simulations.Rmd
  • output Fig2_simulation_results.csv