Dragonfish

Dragonsfish is a functional and taxonomic meta-omic profiler that fully leverages more recent and signficant advances in algorithms and data structures for indexing and querying large-scale genomics data. Dragonfish uses Pufferfish at its core, and therefore a compacted colored de-Bruijn Graph (ccdBG) representation and index. Pufferfish can efficiently index large collections of genome and transcriptome sequences and perform fast and accurate alignment of sequencing reads from genomic and transcriptomic studies, including metagenomic and metatranscriptomic studies.

Dragonfish efficiently and accurately quantifies the abundance of aligned reads to annotated functional features shared across all the genomes and transcriptomes in the index, while also providing strain-, species-, and genus-level taxonomic abundance contributions to functional features. Dragonfish also jointly quantifies taxonomic abundances from mappded reads using Cedar.

By utilizing modern computational methods for indexing and querying large-scale genomic data, we now make it feasible to efficiently index on the order of ~100k full genomic or transcriptomic references, rapidly align reads against an index of such references, efficiently and accurately quantify their abundances, and map and summarize abundances to any annotated functional database features as well as to taxonomy. By accomplishing this, Dragonfish offers significant advantages to existing functional meta-omic profiling tools that are available.

Installation

Mambaforge

Install and set up Miniforge3 Linux:

Project source

Obtain the latest Dragonfish project source code:

git clone https://github.com/hermidalc/dragonfish.git
cd dragonfish

Run Dragonfish

Run the full pipeline:

./scripts/run_snakemake.sh

Additional Notes

Building Pufferfish

I already provide statically built Pufferfish binaries for Linux x86-64. Though here are my notes for trying to build it for other platform and architectures.

I could not get Pufferfish to build within a conda envrionment using the required conda dependencies already installed into that environment. The main issue appears to be with the SeqLib dependency failing to see its required conda dependency C files or having other compilation issues.

Therefore, if you want to make you own Pufferfish build, you have to use dependencies from your system-wide package manager (and therefore potentially need sudo/root permissions), e.g. for RHEL/Fedora Linux dnf or Ubuntu apt. SeqLib successfully builds this way.

Static build

Get the Dragonfish project Pufferfish submodule:

git submodule update --init --recursive

A static build will create fully self-contained binaries for the OS platform and CPU architecture you are using to make the build. These binaries can then simply be copied onto another of the same platform and architecture and will work. On RHEL/Fedora, install the following dependencies using your package manager:

sudo dnf install \
cmake \
curl \
bzip2-static \
gcc \
gcc-c++ \
git \
glibc-static \
hwloc-devel \
jsoncpp-devel \
wget \
xz-static \
zlib-static

Download and build jemalloc:

cd external
https://github.com/jemalloc/jemalloc/archive/refs/tags/5.3.0.zip
unzip 5.3.0.zip
cd jemalloc-5.3.0

./autogen.sh
make
make install

You will find the static library under lib/libjemalloc.a and header under include/jemalloc/jemalloc.h. Now go to the Dragonfish project directory and create the Pufferfish build directory:

mkdir external/pufferfish/build
cd external/pufferfish/build

If your system has CMake >=3.24, then you can build with the following command:

cmake \
-DZLIB_USE_STATIC_LIBS \
-DBZip2_USE_STATIC_LIBS \
-DLZMA_USE_STATIC_LIBS \
-DJEMALLOC_LIBRARY=<path to jemalloc>/lib/libjemalloc.a \
-DJEMALLOC_INCLUDE_DIR=<path to jemalloc>/include \
../

Otherwise you have to do a temporary bit of hacking to prevent the build from including your system compression library dynamic files. In RHEL/Fedora run the following commands:

sudo mv /usr/lib64/libz.so /usr/lib64/libz.so.ORIG
sudo mv /usr/lib64/libbz2.so /usr/lib64/libbz2.so.ORIG
sudo mv /usr/lib64/liblzma.so /usr/lib64/liblzma.so.ORIG

cmake \
-DJEMALLOC_LIBRARY=<path to jemalloc>/lib/libjemalloc.a \
-DJEMALLOC_INCLUDE_DIR=<path to jemalloc>/include \
../

sudo mv /usr/lib64/liblzma.so.ORIG /usr/lib64/liblzma.so
sudo mv /usr/lib64/libbz2.so.ORIG /usr/lib64/libbz2.so
sudo mv /usr/lib64/libz.so.ORIG /usr/lib64/libz.so

Now build with:

make

After the build successfully completes you now have the statically compiled Pufferfish binaries under the build src directory. Check that they are indeed static binaries, for example on Linux run ldd and you should not see any dynamic links to any .so shared objects or linux kernel dynamic linked files:

$ ldd src/*
 not a dynamic executable
src/cedar:
 not a dynamic executable
src/filtersam:
 not a dynamic executable
src/getLineage:
 not a dynamic executable
src/pufferfish:
 not a dynamic executable

Again, these are fully self-contained and can simply be copied and used on any computer with the same OS platform and CPU architecture.

Dynamic build

In RHEL/Fedora, install the following dependencies:

sudo dnf install \
cmake \
curl \
bzip2-devel \
gcc \
gcc-c++ \
git \
glibc-devel \
hwloc-devel \
jsoncpp-devel \
jemalloc-devel \
wget \
xz-devel \
zlib-devel

Instead of using the Dragonfish forked Pufferfish submodule (where I made changes to CMakeLists.txt for static build), get the COMBINE-lab official repository:

git clone [email protected]:COMBINE-lab/pufferfish.git

Build Pufferfish:

mkdir pufferfish/build
cd pufferfish/build
cmake ../
make