1. cluster_identifier/ - this directory contains the part of the application responsible for identifying soft clipped clusters. For how to build see the build section.
2. cluster_analysis/ - contains code related to the analysis of soft-clipped clusters and interpretation.
3. validation/ - sample bam and outputs for testing SCRAMble
4. README - Repository description/notes.

Install dependencies (Ubuntu 20.04):

apt-get update
apt-get install -y  \
    autoconf \
    autogen \
    build-essential \
    curl \
    libbz2-dev \
    libcurl4-openssl-dev \
    libhts-dev \
    liblzma-dev \
    libncurses5-dev \
    libnss-sss \
    libssl-dev \
    libxml2-dev \
    ncbi-blast+ \
    r-base \
    r-bioc-biostrings \
    r-bioc-rsamtools \
    r-cran-biocmanager \
    r-cran-devtools \
    r-cran-stringr \
    r-cran-optparse \
    zlib1g-dev

Install R packages dependencies:

Rscript -e "library(devtools); install_github('mhahsler/rBLAST')"

To build the cluster_identifier (estimated install time <5 minutes):

$ cd cluster_identifier/src
$ make

That should be it. It will create an executable named build/cluster_identifier.

SCRAMble runs as a two-step process. First cluster_identifier is used to generate soft-clipped read cluster consensus sequences. Second, SCRAMble-MEIs.R analyzes the cluster file for likely MEIs. Running SCRAMble on the test bam in the validation directory should take <1 minute for each step.

To run SCRAMble cluster_identifier:

$ /path/to/scramble/cluster_identifier/src/build/cluster_identifier \
    /path/to/install_dir/scramble/validation/test.bam > /path/to/output/test.clusters.txt

To run SCRAMble-MEIs and SCRAMble-dels(with default settings):

$ Rscript --vanilla /path/to/scramble/cluster_analysis/bin/SCRAMble.R \
    --out-name /path/to/output/test 	\
    --cluster-file /path/to/output/test.clusters.txt \
    --install-dir /path/to/scramble/cluster_analysis/bin \
    --mei-refs /path/to/scramble/cluster_analysis/resources/MEI_consensus_seqs.fa \
    --ref /path/to/scramble/validation/test.fa \
    --eval-meis \
    --eval-dels 

SCRAMble is also distributed with a Dockerfile. Running SCRAMble using docker (estimated install time <10 minutes):

$ git clone https://github.com/GeneDx/scramble.git
$ cd scramble
$ docker build -t scramble:latest .
$ docker run -it --rm scramble:latest bash
# cluster_identifier \
    /app/validation/test.bam > /app/validation/test.clusters.txt
# Rscript --vanilla /app/cluster_analysis/bin/SCRAMble.R \
    --out-name ${PWD}/test \
    --cluster-file /app/validation/test.clusters.txt \
    --install-dir /app/cluster_analysis/bin \
    --mei-refs /app/cluster_analysis/resources/MEI_consensus_seqs.fa \
    --ref /app/validation/test.fa \
    --eval-dels \
    --eval-meis

The output of cluster_identifier is a tab delimited text file with clipped cluster consensus sequences. The columns are as follows:

Calling SCRAMble.R with --eval-meis produces a tab delimted file. If a reference .fa file is provided, then a VCF is produced as well. The <out-name>_MEIs.txt output is a tab delimited text file with MEI calls. If no MEIs are present an output file will still be produced with only the header. The columns are as follows:

Calling SCRAMble.R with --eval-dels produced a VCF and a tab delimted file. The <out-name>_PredictedDeletions.txt output is a tab delimited text file with deletion calls. If no deletions are present an output file will still be produced with only the header. The columns are as follows:

In theory, SCRAMble should work well on any MEI reference fasta sequences, however, it has only been tested on the sequences provided in /path/to/scramble/cluster_analysis/resources/MEI_consensus_seqs.fa.