/diamond

Accelerated BLAST compatible local sequence aligner.

Primary LanguageC

DIAMOND v0.7.9 by Benjamin Buchfink - http://github.com/bbuchfink/diamond

Published in Nature Methods 12, 59–60 (2015) | doi:10.1038/nmeth.3176

DIAMOND is a BLAST-compatible local aligner for mapping protein and translated DNA query sequences against a protein reference database (BLASTP and BLASTX alignment mode). The speedup over BLAST is up to 20,000 on short reads at a typical sensitivity of 90-99% relative to BLAST depending on the data and settings.

Download & Installation

If you use a recent Linux operating system, you can download the software in binary format for immediate use:

wget http://github.com/bbuchfink/diamond/releases/download/v0.7.9/diamond-linux64.tar.gz
tar xzf diamond-linux64.tar.gz

Users of Mac OS X and some old Linux systems need to compile the software from source (see Compiling from source).

Basic command line use

We assume to have a protein database file in FASTA format named nr.faa and a file of DNA reads that we want to align named reads.fna.

In order to set up a reference database for DIAMOND, the makedb command needs to be executed with the following command line:

$ diamond makedb --in nr.faa -d nr

This will create a binary DIAMOND database file with the specified name (nr.dmnd). The alignment task may then be initiated using the blastx command like this:

$ diamond blastx -d nr -q reads.fna -a matches -t <temporary directory>

The temporary directory should point to a fast local disk with a lot of free space. It is possible to omit this option, this will however increase the program's memory usage substantially.

The output file here is specified with the –a option and named matches.daa. It is generated in DAA (DIAMOND alignment archive) format. Other formats can be generated using the view command. For instance, the following command will generate BLAST tabular format from the DAA file and save it to disk:

$ diamond view -a matches.daa -o matches.m8

Commands

Commands are issued as the first parameter on the command line and set the task to be run by the program.

Command Description
makedb Create DIAMOND formatted reference database from a FASTA input file.
blastp Align protein query sequences against a protein reference database.
blastx Align translated DNA query sequences against a protein reference database.
view Generate formatted output from DAA files.

Makedb options

Option Short Default Description
--threads -p max Number of CPU threads.
--in     Path to protein reference database file in FASTA format (may be gzip compressed).
--db -d   Path to DIAMOND database file.
--block-size -b 2 Block size in billions of sequence letters to be processed at a time.

General & IO options

Option Short Default Description
--threads -p max Number of CPU threads.
--db -d   Path to DIAMOND database file (not including the file extension .dmnd).
--query -q   Path to query input file in FASTA or FASTQ format (may be gzip compressed).
--daa -a   Path to output file in DAA format (extension .daa will be appended).

Sensitivity & speed options

Option Short Default Description
--sensitive     Trigger the sensitive alignment mode with a 16x9 seed shape configuration.
--band   auto Dynamic programming band for seed extension. This corresponds to the maximum length of gaps that can be found in alignments.

Scoring & Reporting Options

Option Short Default Description
--gapopen   11 Gap open penalty.
--gapextend   1 Gap extension penalty.
--matrix   BLOSUM62 Scoring matrix.
--seg     Enable SEG masking of low complexity segments in the query (yes/no). The default is no for blastp and yes for blastx.
--max-target-seqs -k 25 The maximum number of target sequences per query to keep alignments for.
--top     Keep alignments within the given percentage range of the top alignment score for a query (overrides –max-target-seqs option).
--evalue -e 0.001 Maximum expected value to keep an alignment.
--min-score     Minimum bit score to keep an alignment. Setting this option will override the --evalue parameter.

Memory & performance options

Option Short Default Description
--tmpdir -t /dev/shm Directory to be used for temporary storage.
--index-chunks -c 4 The number of chunks for processing the seed index.

It is recommended to always use the --tmpdir option and set this to a disk-based directory. The amount of disk space that will be used depends on the program's settings and your data. As a general rule you should ensure that 100 GB of disk space are available here. If you run the program in a cluster environment, and disk space is only available over a slow network based file system, you may want to omit the --tmpdir option. This will keep temporary information in memory and increase the program's memory usage substantially.

The --block-size/-b option is set for the makedb command and the main option for controlling the program's memory usage. Bigger numbers will increase the use of memory and temporary disk space, but also improve performance. The program can be expected to roughly use six times this number of memory (in GB). So for the default value of -b=2, the memory usage will be about 12 GB.

The --index-chunks/-c option can be additionally used to tune the performance. It is recommended to set this to 1 on a high memory server, which will increase performance and memory usage, but not the usage of temporary disk space.

View options

Option Short Default Description
--daa -a   Path to input file in DAA format.
--out -o   Path to output file.
--outfmt -f   Format of output file. (tab = BLAST tabular format; sam = SAM format)
--compress   0 Compression for output file (0=none, 1=gzip).

FAQ

DIAMOND is slower than claimed in the paper, even slower than BLAST.

The DIAMOND algorithm is designed for the alignment of large datasets. The algorithm is not efficient for a small number of query sequences or only a single one of them, and speed will be low. BLAST is recommend for small datasets.

Can several copies of DIAMOND be run in parallel?

It is possible, but not recommended. The algorithm is more efficient if you allocate more memory to a single task. If you need to process several files, performance will be better if you run DIAMOND on them sequentially.

Reads imported into MEGAN lack taxonomic or functional assignment.

MEGAN requires mapping files which need to be downloaded separately at the MEGAN website and configured to be used.

Compiling from source

The requirements for compiling DIAMOND are Boost (version 1.53.0 or higher) and zlib. If a system-wide Boost installation is not possible, the package includes a script called install-boost which will download and install a local copy of Boost for the user.

To compile DIAMOND from source, invoke the following commands on the shell:

$ wget http://github.com/bbuchfink/diamond/archive/v0.7.9.tar.gz
$ tar xzf v0.7.9.tar.gz
$ cd diamond-0.7.9/src
$ ./install-boost                                       # optional, for installing Boost
$ make

The diamond binary will be created in diamond-0.7.9/bin.

Compiling using CMake

To compile DIAMOND from source, invoke the following commands on the shell:

$ wget http://github.com/bbuchfink/diamond/archive/v0.7.9.tar.gz
$ tar xzf v0.7.9.tar.gz
$ cd diamond-0.7.9
$ mkdir build
$ cd build
$ cmake .. # Use cmake -DCMAKE_INSTALL_PREFIX=... to install to a different prefix.
$ make install

Installing using Homebrew/Linuxbrew

$ brew install homebrew/science/diamond

Scoring matrices

Matrix Supported values for (gap open)/(gap extend)
BLOSUM45 (10-13)/3; (12-16)/2; (16-19)/1
BLOSUM50 (9-13)/3; (12-16)/2; (15-19)/1
BLOSUM62 (6-11)/2; (9-13)/1
BLOSUM80 (6-9)/2; 13/2; 25/2; (9-11)/1
BLOSUM90 (6-9)/2; (9-11)/1
PAM250 (11-15)/3; (13-17)/2; (17-21)/1
PAM70 (6-8)/2; (9-11)/1
PAM30 (5-7)/2; (8-10)/1