Ray is a parallel de novo genome assembler that utilises the message-passing interface everywhere and is implemented using peer-to-peer communication.
Ray is free software distributed under the terms of the GNU General Public License, version 3 (GPLv3).
Ray is implemented using RayPlatform, a message-passing-interface programming framework.
Ray is documented in
- Documentation/ (many files)
- MANUAL_PAGE.txt (command-line options, same as Ray -help)
- README.md (general)
- INSTALL.txt (quick installation)
Standard:
- de novo genome assembly (works by default) => http://online.liebertpub.com/doi/abs/10.1089/cmb.2009.0238 => Documentation/README-heuristics
- quantification of contig abundances (works by default)
Metagenomics:
- Ray Meta: de novo metagenome assembly (works by default) => http://genomebiology.com/2012/13/12/R122
- Ray Communities: quantification of microbiome consortia members (with Ray Communities with -search) => Documentation/BiologicalAbundances.txt
- Ray Communities: taxonomy profiling of samples (with -search and -with-taxonomy) => Documentation/Taxonomy.txt
- Ray Ontology: gene ontology profiling of samples (with -search and -gene-ontology) => Documentation/GeneOntology.txt
- Ray Surveyor: compare genomic content between samples (with -run-surveyor) => Documentation/Ray-Surveyor.md
Transcriptomics:
- de novo transcriptome assembly (works, but not tested a lot)
- quantification of transcript expression
- Geeknet, Inc. http://sourceforge.net/projects/denovoassembler/
- GitHub, Inc. https://github.com/sebhtml/ray
- Debian (Software in the Public Interest, Inc.) http://packages.debian.org/sid/ray
- Ubuntu (Canonical Ltd.) https://launchpad.net/ubuntu/+source/ray
- archlinux (AUR Development Team.) https://aur.archlinux.org/packages/ray/
- DNAnexus, Inc. https://platform.dnanexus.com/app/ray
- CloudBioLinux https://github.com/chapmanb/cloudbiolinux/blob/master/cloudbio/custom/bio_nextgen.py
In progress:
- Fedora (Red Hat, Inc.) https://bugzilla.redhat.com/show_bug.cgi?id=872783 (in progress)
- Galaxy (Galaxy Team) http://user.list.galaxyproject.org/How-do-I-add-Ray-to-Galaxy-Central-in-the-tool-shed-td4655623.html#none
- BaseSpace (Illumina, Inc.)
-
http://github.com/sebhtml/ray (Ray plugins for genomics)
-
http://github.com/sebhtml/RayPlatform (the engine RayPlatform)
If you want to contribute, clone the repository, make changes and I (Sébastien Boisvert) will pull from you after reviewing the code changes.
-
http://github.com/sebhtml/Ray-TestSuite (system tests & unit tests)
-
http://github.com/sebhtml/Ray.web (Ray SourceForge web site)
-
http://github.com/sebhtml/ray-logo (artworks)
-
Users: denovoassembler-users AT lists.sourceforge.net
-
Read it on gmane: http://blog.gmane.org/gmane.science.biology.ray-genome-assembler
-
Development/hacking: denovoassembler-devel AT lists.sourceforge.net
-
SEQanswers: http://seqanswers.com/forums/showthread.php?t=4301
You need a C++ compiler (supporting C++ 1998), make, an implementation of MPI (supporting MPI 2.2).
tar xjf Ray-x.y.z.tar.bz2
cd Ray-x.y.z
make PREFIX=build
make install
ls build
tar xjf Ray-x.y.z.tar.bz2
cd Ray-x.y.z
mkdir build
cd build
cmake ..
make
make PREFIX=build2000 MPICXX=/software/openmpi-1.4.3/bin/mpicxx
make install
Tested C++ compilers: see Documentation/COMPILERS.txt
To compile with MPI I/O, use this:
make MPI_IO=y
Some processors have the popcnt instruction and other cool instructions. With gcc, add -march=native to build Ray for the processor used for the compilation.
make PREFIX=Build.native DEBUG=n ASSERT=n EXTRA=" -march=native"
make install
Another way to build Ray is to use whole-program optimization. With gcc, use this script:
./scripts/Build-Link-Time-Optimization.sh
make PREFIX=Ray-Large-k-mers MAXKMERLENGTH=64
# wait
make install
mpirun -np 512 Ray-Large-k-mers/Ray -k 63 -p lib1_1.fastq lib1_2.fastq \
-p lib2_1.fastq lib2_2.fastq -o DeadlyBug,Assembler=Ray,K=63
# wait
ls DeadlyBug,Assembler=Ray,K=63/Scaffolds.fasta
make PREFIX=build-3000 MAXKMERLENGTH=64 HAVE_LIBZ=y HAVE_LIBBZ2=y \
ASSERT=n FORCE_PACKING=y
# wait
make install
ls build-3000
see the Makefile for more.
To run Ray on paired reads:
mpiexec -n 25 Ray -k31 -p lib1.left.fasta lib1.right.fasta -p lib2.left.fasta lib2.right.fasta -o RayOutput
ls RayOutput/Contigs.fasta
ls RayOutput/Scaffolds.fasta
ls RayOutput/
Ray can be run with a configuration file instead.
mpiexec -n 16 Ray Ray.conf
Content of Ray.conf:
-k 31 # this is a comment -p lib1.left.fasta lib1.right.fasta
-p lib2.left.fasta lib2.right.fasta
-o RayOutput
RayOutput/Contigs.fasta and RayOutput/Scaffolds.fasta
type Ray -help for a full list of options and outputs
Ray assembles color-space reads and generate color-space contigs. Files must have the .csfasta extension. Nucleotide reads can not be mixed with color-space reads. This is an experimental feature.
- Assembler: https://www.ncbi.nlm.nih.gov/pubmed/20958248
- Profiler: https://www.ncbi.nlm.nih.gov/pubmed/23259615
- Surveyor: https://www.ncbi.nlm.nih.gov/pubmed/28957508
cd code
doxygen DoxygenConfigurationFile
cd DoxygenDocumentation/html
firefox index.html
- http://aws.amazon.com/ec2/hpc-applications/
- https://cloud.genomics.cn/
- http://szdaily.sznews.com/html/2011-08/04/content_1689998.htm
- http://www.nature.com/nbt/journal/v28/n7/full/nbt0710-691.html
- http://dskernel.blogspot.com/2011/07/understand-main-loop-of-message-passing.html
- http://cw.squyres.com/
- http://blogs.cisco.com/performance/
- http://www.parawiki.org/index.php/Message_Passing_Interface#Peer_to_Peer_Communication
- https://en.wikipedia.org/wiki/Actor_model
- https://arxiv.org/abs/1008.1459
- http://dl.acm.org/citation.cfm?id=1624804
- Doctoral Award to S.B., Canadian Institutes of Health Research (CIHR)
- Doctoral Award to M.D., The Fonds de recherche du Québec – Santé (FRQS)
see AUTHORS
see Documentation/VISUAL_STUDIO.txt