GSE220185 Processing Pipeline

Publication

Molecular cell (2023) — PMID 37084731

Dataset

Proteomic discovery of chemical probes that perturb protein complexes in human cells

1. 1

   Processed using Skipper https://github.com/YeoLab/skipper

   Skipper vNot specified GitHub

   $ Bash example

   # Installation of Snakemake and cloning Skipper workflow
   # It is recommended to use a dedicated conda environment for Snakemake and its dependencies.
   # conda create -n skipper_env snakemake mamba -y
   # conda activate skipper_env
   
   # Clone the Skipper workflow repository
   # git clone https://github.com/YeoLab/skipper.git
   # cd skipper
   
   # The Skipper workflow uses conda environments defined within its rules.
   # These environments will be created automatically by Snakemake if --use-conda is specified.
   
   # Execution of Skipper workflow
   # This is a generic command. Actual parameters (input files, genome, etc.)
   # would be specified in a configuration file (e.g., config/config.yaml).
   #
   # Reference dataset: For eCLIP, a genome assembly like hg38 or mm10 is typically required.
   # This would be specified in the config.yaml (e.g., genome: hg38).
   #
   # Replace 'N' with the desired number of CPU cores.
   # Replace 'path/to/your/config.yaml' with the path to your specific configuration file.
   # Replace 'path/to/your/output_directory' with your desired output location.
   # A minimal config.yaml might look like:
   # samples:
   #   sample1:
   #     R1: "path/to/sample1_R1.fastq.gz"
   #     R2: "path/to/sample1_R2.fastq.gz" # if paired-end
   # genome: "hg38" # or "mm10", etc.
   # annotation: "path/to/your/gtf_or_gff"
   #
   # For a full run, you would typically copy and modify the example config file from the Skipper repository.
   
   snakemake --snakefile Snakefile --cores N --use-conda --configfile path/to/your/config.yaml --directory path/to/your/output_directory

   Copy View on GitHub
2. 2

   Skipper trims the read, extract UMI, align to the genome and deduplicate

   Skipper vSnakemake workflow (dependencies: Snakemake 6.0.5, cutadapt 3.4, STAR 2.7.9a, UMI-tools 1.1.2, Picard 2.25.7) GitHub

   $ Bash example

   # Install Git and Conda if not already present
   # sudo apt-get update && sudo apt-get install -y git
   # wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda.sh
   # bash miniconda.sh -b -p $HOME/miniconda
   # export PATH="$HOME/miniconda/bin:$PATH"
   # conda init bash
   # source ~/.bashrc
   
   # Clone the Skipper pipeline repository
   # git clone https://github.com/yeolab/skipper.git
   # cd skipper
   
   # Create and activate the conda environment for the pipeline
   # conda env create -f environment.yaml
   # conda activate skipper_env
   
   # --- Configuration for Skipper ---
   # Create a config.yaml file with your specific parameters.
   # Replace placeholder paths and values with your actual data.
   cat << EOF > config.yaml
   GENOME_DIR: "/path/to/STAR_index/hg38" # Path to STAR genome index directory (e.g., built from hg38)
   GENOME_FASTA: "/path/to/genome/hg38.fa" # Path to the genome FASTA file (e.g., hg38.fa)
   GTF: "/path/to/annotations/gencode.v38.annotation.gtf" # Path to the gene annotation GTF file (e.g., gencode.v38.annotation.gtf)
   
   ADAPTER_FWD: "AGATCGGAAGAGCACACGTCTGAACTCCAGTCA" # Forward adapter sequence for trimming
   ADAPTER_REV: "AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT" # Reverse adapter sequence for trimming
   
   UMI_PATTERN: "NNNNNNNN" # UMI pattern (e.g., 8 N's for an 8-bp UMI)
   # BARCODE_PATTERN: "NNNNNN" # Optional: Barcode pattern if present, otherwise comment out or leave empty
   
   DEDUP_METHOD: "directional" # Deduplication method: "directional", "unique", or "position"
   
   # Define your samples and input directory
   SAMPLES: ["sample1", "sample2"] # List of sample names (e.g., corresponding to sample1_R1.fastq.gz, sample2_R1.fastq.gz)
   INPUT_DIR: "/path/to/raw_fastq" # Directory containing your raw FASTQ files
   OUTPUT_DIR: "results" # Directory for output files
   EOF
   
   # --- Run the Skipper pipeline ---
   # This command executes the Snakemake workflow using the specified configuration.
   # Adjust --cores based on available CPU resources.
   snakemake --use-conda --cores 8 --configfile config.yaml all

   Copy View on GitHub
3. 3

   To find enriched windows, it models the CLIP libraries crosslinking-induced truncation(CITs) with beta-binomial model with GC-bias of each window as a varaible

   PureCLIP (Inferred with models/gemini-2.5-flash) v1.3.1

   $ Bash example

   # Install PureCLIP via Bioconda
   # conda install -c bioconda pureclip
   
   # Example command to find enriched windows using PureCLIP
   # Replace <input_clip_replicateX.bam> with your aligned CLIP-seq BAM files,
   # <genome.fa> with your reference genome FASTA file (e.g., hg38.fa),
   # and <output_enriched_windows.bed> with your desired output file name.
   # Adjust -nt for the number of threads to use.
   
   PureCLIP -i input_clip_replicate1.bam input_clip_replicate2.bam -g hg38.fa -o output_enriched_windows.bed -nt 8

   Copy
4. 4

   overdispersion parameters are estimated from two input replicates

   DESeq2 (Inferred with models/gemini-2.5-flash) v1.42.0 GitHub

   $ Bash example

   # Install R and DESeq2 if not already available
   # conda create -n deseq2_env r-base r-essentials bioconductor-deseq2
   # conda activate deseq2_env
   
   # Example R script to estimate overdispersion parameters using DESeq2
   # This script assumes a count matrix and a sample information file as input.
   
   cat << 'EOF' > estimate_overdispersion.R
   library(DESeq2)
   
   # --- User-defined parameters ---
   count_matrix_file <- "counts.tsv" # Combined count matrix from replicates
   sample_info_file <- "sample_info.csv" # Sample metadata
   
   output_dispersion_file <- "overdispersion_estimates.tsv"
   # -------------------------------
   
   # Load count data
   count_data <- read.delim(count_matrix_file, row.names = 1, check.names = FALSE)
   # Ensure counts are integers
   count_data <- round(count_data)
   count_data[is.na(count_data)] <- 0
   count_data <- as.matrix(count_data)
   
   # Load sample information
   sample_info <- read.csv(sample_info_file, row.names = 1)
   
   # Ensure sample names match and are in the same order
   sample_info <- sample_info[colnames(count_data), , drop = FALSE]
   
   # Create DESeqDataSet object
   # For simple overdispersion estimation, a minimal design is sufficient.
   # If comparing two replicates, the design might be ~1 or ~replicate_group
   dds <- DESeqDataSetFromMatrix(countData = count_data,
                                 colData = sample_info,
                                 design = ~ 1) # A simple design to estimate dispersions
   
   # Run DESeq2 pipeline (this includes dispersion estimation)
   dds <- estimateSizeFactors(dds)
   dds <- estimateDispersions(dds) # Explicitly estimate dispersions
   
   # Extract dispersion estimates
   dispersions <- mcols(dds)$dispersion
   names(dispersions) <- rownames(dds)
   
   # Save dispersions
   write.table(as.data.frame(dispersions), file = output_dispersion_file,
               sep = "\t", quote = FALSE, col.names = NA)
   
   message(paste("Overdispersion parameters estimated and saved to:", output_dispersion_file))
   EOF
   
   # Create dummy count data (replace with actual data)
   echo -e "gene\trep1\trep2" > counts.tsv
   echo -e "geneA\t100\t120" >> counts.tsv
   echo -e "geneB\t50\t65" >> counts.tsv
   echo -e "geneC\t200\t210" >> counts.tsv
   echo -e "geneD\t10\t15" >> counts.tsv
   
   # Create dummy sample info (replace with actual data)
   echo -e "sample,condition" > sample_info.csv
   echo -e "rep1,control" >> sample_info.csv
   echo -e "rep2,control" >> sample_info.csv
   
   # Execute the R script
   Rscript estimate_overdispersion.R

   Copy View on GitHub

Tools Used