Command-line tumor-calling & data analysis tools for tumor size profiling by Tumor-Barcoded ultra-deep sequencing. Please refer to our publication in Nature Methods for details.
This package includes a pipeline for processing Illumina ® Single-end or Paired-end reads (Paired-end is recommended) into unique barcode pileups and their size in Absolute Cell Number, as described in our original Tuba-Seq paper. Additional tools are provided, including:
- Identification of contaminating DNA and a BLAST-search based investigation of the source of contaminating DNA;
- Analysis of the sequencing error-rate of the Illumina run;
- Analysis of barcode randomness, total diversity of the viral library, and evidence for cross-contamination within the library;
- Statistical tools to summarize tumor size distributions and mice cohorts--including identification of outliers;
- Graphing functions to illustrate size distributions.
This pipeline uses the DADA2 de-noising and sample inference algorithm to identify unique DNA barcodes. Because tumors differ in size by >1,000-fold, it is difficult to delineate small tumors from recurrent read errors (this issue has been well-described previously). As such, DADA2 statistical modeling is computationally-intesnive and tools for broadcasting this step onto a distributed computing cluster are provided. Sequencing error rates are estimated from the non-degenerate regions of DNA barcodes using a novel approach. The pipeline is also extensible to other barcode clustering algorithms, including direct support for Bartender, which clusters barcodes in a fraction of the time. However, in our experience DADA2 is necessary for faithful tumor calling.
Proficiency with python, its scientific computing modules, command-line scripting, distributed clusters, and patience is recommended.
Installation from source is intended for POSIX-style Operating Systems (e.g. Linux, darwin). The prequisites are:
- Python 3.2 or later
- R 3.2 or later
- gcc or clang
Executable scripts, by virtue of their shebang, assume python3 and R are within the user's $PATH. The following commands will generally complete installation:
# Download Source Code & Submodules
git clone --recursive https://github.com/petrov-lab/tuba-seq.git
cd tuba-seq
# Compile the Striped Smith-Waterman Read Alignment Library
cd Complete-Striped-Smith-Waterman-Library/src
make
cd ../..
# Install the tuba-seq Python package--use pip to install requirements.
pip3 install .
# Install DADA2 from source
./INSTALL_DADA2.R
# Add command-line tools to User's Path (optional)
cp bin/* ~/binTo merge paired-end reads, I recommend using the PEAR Algorithm (installation instructions here). This pipeline is in a very nascent state, so there is no troubleshooting manual. Please contact me, if you cannot troubleshoot install problems yourself.
Feedback is most appreciated. Please contact me at cmcfarl2 [at] stanford [dot] edu with any questions or suggestions.
Documentation and source code can be found at https://github.com/petrov-lab/tuba-seq
Contains all executables. Python script usage can always be accessed via the --help command, e.g.
bin/preprocess.py --help
R scripts are self-documented and several very-simple shell scripts are provided to illustrate the typical pipeline workflow. They're best understood by inspection.
Contains the version of this code that was used to generate the data published in Rodgers et al. (2017). Behavior of this version remains similar to the modern version, but it is not recommended for use (other than for reproducing our published findings). This version lacks considerable functionality and does not incorporate what we believe to be the, evolving, best-practices.
Test suite.
Directory of shared python modules.