When a user wants to customize a statistical test or has to handle a low-quality dataset, the standalone version of CRISPRcloud pipeline will be a good option to deal with that requirements. This pipeline is built with Snakemake (Köster and Rahmann, 2012), and it can be easily customized by a user. We describe how to use the pipeline and show a way to add a custom test in this section.

1. Installation

The pipeline has tested on MacOS and Ubuntu successfully, and the following applications must be installed before using it.

• Git: https://git-scm.com/
• R: https://cran.r-project.org/
• Python3: https://www.python.org/download/releases/3.0/
• Snakemake: http://snakemake.readthedocs.io/en/stable/

First, run the below command on the terminal. A sudo/root account will need to install additional R packages.

$ git clone https://github.com/hyunhwaj/CRISPRcloud-standalone
$ ./install_packages.sh

2.Preparing dataset

There are several files which user need to prepare to run the pipeline.

• CRISPR pooled screening RNAseq data: fastq files(or compressed fastq file), recommends to locate "reads" folder.
• sgRNA library annotations: a fasta file, in the fasta file of your library, a name of each sgRNA should be formatted as >genename_<index>, and <index> should be an integer. Otherwise, CRISPRcloud cannot handle your library. An example of the file is located in "lib" folder (yusa_library.fa or ysa_library_small.fa), and a git repository https://github.com/hyunhwaj/GeCKOv2_fasta contains GeCKO v2 knockout library for human and mouse.
• Label for each RNAseq data: an example is stored at data folder as label.txt. Each row represents information of a sample. The first column in the row should be the base name of an RNAseq file without extension (i.e. CRISPR_SAMPLE_1 should be a value of the column if the filename is CRISPR_SAMPLE_1.fastq or CRISPR_SAMPLE_1.fastq.gz). The next column should contain a name of a group of the sample, and the value of the last column should be an integer and represent the index number of its replicate. If the same index number appear multiple times in a group, then the corresponded samples are considered as a one replicates in the pipeline.

3. Edit configurations in Snakemake

In the Snakemake file in the root folder, there are a couple of parameters user should set-up.

In rule quant_samples

parameters	Contents
input.lib	The path of the sgRNA library
input.fq	A wildcard path of the RNAseq data
params.adapt	The 5' adaptor sequence which should be trimmed in a RNAseq
params.err_trim	Parameter for accepting mismatch for trimming. cutadapt is used for the process
params.mismatch	Parameter for accepting mismatch for the sgRNA quantification

In run_analysis

parameters	Contents
params.lib_path	The path contains all scripts in the pipeline
params.input_path	The path will contain inputs such as label and data summarization
params.output_path	The path will contain output files
params.study	Type of design of the study (dropout or enrichment)
params.base_group	The baseline group in the pooled study experiment
params.stat_test	The hypothesis test method will be used in the pipeline (logt, deseq2 or ibb)

4. Run pipeline

When the entire setup is done, then run the following command to start the Snakemake pipeline.

$ Snakemake

Snakemake also allows multi-core processing with the -ncores parameter.

$ Snakemake –ncores=4

5. Add a customized test

For a customized test, user must add a function and modified bin/stat.test.R with the following instruction:

1. add a function like below, the parameters are the read counts for each group to run the test.

run.deseq2 <- function(cnt.t0, cnt.t1) {
  library(DESeq2)
  ret <- list()
  col.data <- data.frame(condition=rep("T0", ncol(cnt.t0)), row.names = colnames(cnt.t0))
  col.data <- rbind(col.data, data.frame(condition=rep("T1", ncol(cnt.t1)), row.names=colnames(cnt.t1)))
  col.data$condition <- factor(col.data$condition, levels = c("T0", "T1"))
  print(col.data)
  count.matrix <- cbind(cnt.t0, cnt.t1)
  ckCDS <- DESeqDataSetFromMatrix(countData = round(count.matrix),
                                colData = col.data,
                                design =~condition)
  dds <-DESeq(ckCDS)
  res <- as.data.frame(results(dds))
  ret$fc <- res$log2FoldChange
  ret$p.value <- res$pvalue
  ret$p.adjust <- res$padj
  return(ret)
}

2. If the function is run.custom and if parameter.stat_test in Snakemake would be custom then put the code between line 119 and line 120 like this:

    else if(sel.test=="custom") {
      ret <- run.deseq2(cnt.t0, cnt.t1)
    }