clubcpg_downsample

Using downsample to account for errors caused by read depth differences among samples NOTE: every step is performed on CluBCpG output on one chromosome NOTE: all pathname used in the scripts are absolute path

• Downsample procedure flowchart
• The procedure uses output from clubcpg-cluster command in the package CluBCpG. For more information about CluBCpG, you can read the documentation here
• The three blue blocks represent the three scripts that are used in the procedure. And the yellow block represent the procedure of downsampling and the requierd files.

1. Create cluster patterns

• The cluster patterns in the clubcpg_cluster output are not consistent among different bins. So before performing downsampling, the user need to use this script to create cluster pattern labels that are consistent.
• Inputs:
  • --samples: a list of pathnames to clubcpg-cluster output files (.csv),
  • -o: the output directory
  • --name: the file name (start with "/" and ends with ".csv")
  • NOTE: The list of path names of clubcpg outputs can be of any length, so that the user can put as many clubcpg-cluster output as they want to make the labels consistent across multiple analysis.
• Output: a .csv file containing the labels to the methylation patterns found in the given list of clubcpg-cluster outputs.

Sample command:

 python get_cluster_pattern.py --samples "clubcpg output directory 1" "clubcpg output directory 2" -o "output directory" --name "/filename.csv"

2. Create lowest common read depth

• Create a .csv file summarizing the lowest read depth seen in each bin

• Inputs:

  • --samples: a list of pathnames to clubcpg-cluster output files (.csv),
  • -o: the output directory
  • --name: the file name (start with "/" and ends with ".csv")
  • NOTE: The list of path names of clubcpg outputs can be of any length, so that the user can put as many clubcpg-cluster output as they want to make the labels consistent across multiple analysis.

• Output: a .csv file containing the lowest commmon read depth of each bin in the given list of clubcpg-cluster outputs.

  Sample command:

   python get_lc_reads.py --samples "clubcpg output directory 1" "clubcpg output directory 2" -o "output directory" --name "/filename.csv"
  

3. Perform downsampling procedure

• use the previously generated two files and clubcpg output to perform downsampling. You can run python downsample.py --help to see the commands
  • required input:
    • pathname to the created cluster pattern file (.csv file)
    • pathname to the created lowest common read depth file (.csv file).
    • -A : pathname to a selected clubcpg-cluster output that contains the sample of interest for downsampling.
    • If a user want to perform downsampling on two samples from 2 different clubcpg outputs (e.g.: downsample on sample A from clubcpg output 1 and sample B from clubcpg output 2), they have the following optional arguments:
      • -B :pathname to the 2nd clubcpg-cluster output file that contains the 2nd sample of interest for downsampling.
      • -sampleA: identify which sample to choose from clubcpg output 1 (-A), allowed inputs are "A" or "B"
      • -sampleB: identify which sample to choose from clubcpg output 2 (-B), allowed inputs are "A" or "B"
      • NOTE: to perform downsampling on two samples from 2 different clubcpg outputs, the user must use the three arguments above.
      • For more information on how to use 2 file mode, see section 4.
  • recommended input:
    • -ncore: the number of cpu cores that will be used for running this script. Using more cores will speed up the process.
    • -o : the output directory
    • -name: the name of the output file (start with "" and ends with .csv) Sample command:

      python downsample.py "cluster_patterns path" "lowest common read depths path" -A "clubcpg output path" -B "clubcpg output path 2" -ncore 10 -o "output directory path" -name "/output2.csv"
    

4. Example for using downsampling:

• Background: Assume we want to compare male, neuron, p12 (12 days after birth) data and male,neuron,p35 data.
  • If what we have is clubcpg-cluster output files from an analysis of the two desired samples (i.e. a cluster output file comparing reads from male, neuron, p12 (12 days after birth) data and male,neuron,p35 data
    • Then we only need one sample, so the arguments needed is only -A
  • If the desired samples are in two separate clubcpg-cluster output data, then we need to use 2 file mode;
    • Assume we have 2 cluster files:
      1. clubcpg-cluster output 1 on male, neuron, p12 and female neuron p12, where in the "class split" column, A male and Bfemale
      2. clubcpg-cluster output 2 on male, neuron, p35 and female neuron p35, where in the "class split" column, A male and Bfemale
    • Then we need the following arguments:
      • -A : pathname to clubcpg-cluster output 1
      • -B : pathname to clubcpg-cluster output 2
      • --sampleA "A"
      • --sampleB "B"
• Summary of example
•