Test_DEG_methods-single_cell

Container and scripts to test different differential gene expression analysis methods

Several Differential Gene Expression analsyis methods exhist, single cell RNA-Seq bring new challenges to correctly identify the differentially expressed genes. This repo contains containers, scripts and notebook to run the analysis and check the results:

• Container: vergaju/test_deg contains
  • Seurat
  • DESeq2
  • limma
  • edgeR
  • RankProd (wilcoxon test)
  • sva/combat: batch correction
  • ClusterProfiler and GOsemsim: pathway enrichment analysis
  • other packages for visualization and data wrangling
• Rmd files:
  • Seurat
  • Run pseudobulk methods
  • Rmd to compile the report
• Python env:
  • Scanpy
  • pandas: create pseudobulks
• additional scripts:
  • convert Rds to H5ad and vice versa

Usage:

Starting from an Rds file or a H5ad:

• Use the scripts to convert the files in the desired format
• Obtain the pseudobulks

Single cell methods

Use directly the Rds or H5ad files with scanpy or seurat

Pseudobulk

Use the outputs from the notebook to obtain the pseudobulks as inputs of the DEG methods (pseudobulk = counts, esperimental desing = design matrix)

Container:

Clone the repo and buil the container or

docker pull vergaju/test_deg

Usage:

docker run --rm -p 8888:8787 -e PASSWORD=pass -v $(pwd):/home/rstudio vergaju/test_deg

Open your browser at the address: 'localhost:8888', login with:

• username: rstudio
• password: pass (or select a preferred password)

NOTE: The container should be run in the same folder with the Rmd files#

Mindmap:

    flowchart LR
      subgraph Prepare_files
      A[Start from Rds] --> B{convert scripts}
      C[Start from H5ad] --> B
      B --> D[H5ad]
      B --> E[Rds]
      D --> F[Pseudobulk and desing matrix]
      end
      subgraph scDEG
      D --> G([scanpy DEG])
      E --> H([Seurat DEG])
      end
      subgraph pseudobulk
      F --> I([DESeq2 with or without batch correction])
      F --> J([limma with or without batch correction])
      F --> K([edgeR with or without batch correction])
      F --> L([RankProd or wilcoxon with or without batch correction])
      end
      subgraph reporting
      G --> M([Report])
      H --> M
      I --> M
      J --> M
      K --> M
      L --> M
      M --> N([Decide best method])
      end

Loading