sylph is a program that performs ultrafast (1) ANI querying or (2) metagenomic profiling for metagenomic shotgun samples.
Containment ANI querying: sylph can search a genome, e.g. E. coli, against your sample. If sylph outputs an estimate of 97% ANI, your sample contains an E. coli with 97% ANI to the queried genome.
Metagenomic profiling: sylph can determine the species/taxa in your sample and their abundances, just like Kraken or MetaPhlAn.
Profiling 1 Gbp of mouse gut reads against 85,205 genomes in a few seconds
-
Precise species-level profiling: Our tests show that sylph is more precise than Kraken and about as precise and sensitive as marker gene methods (MetaPhlAn, mOTUs).
-
Ultrafast, multithreaded, multi-sample: sylph can be > 50x faster than MetaPhlAn for multi-sample processing. sylph only takes ~15GB of RAM for profiling against the entire GTDB-R220 database (110k genomes).
-
Accurate (containment) ANIs down to 0.1x effective coverage: for bacterial ANI queries of > 90% ANI, sylph can often give accurate ANI estimates down to 0.1x coverage.
-
Easily customized databases: sylph can profile against metagenome-assembled genomes (MAGs), viruses, eukaryotes, and more. Taxonomic information can be incorporated downstream for traditional profiling reports.
sylph uses a k-mer containment method, similar to sourmash or Mash. sylph's novelty lies in using a statistical technique to correct ANI for low coverage genomes within the sample, giving accurate results for low abundance genomes. See here for more information on what sylph can and can not do.
Profile metagenome sample against GTDB-R220 (113,104 bacterial/archaeal species representative genomes)
# download GTDB-R220 pre-built database (~13 GB)
wget https://storage.googleapis.com/sylph-stuff/gtdb-r220-c200-dbv1.syldb
# multi-sample paired-end profiling (sylph version >= 0.6)
sylph profile gtdb-r220-c200-dbv1.syldb -1 *_1.fastq.gz -2 *_2.fastq.gz -t (threads) > profiling.tsv
# multi-sample single-end profiling
sylph profile gtdb-r220-c200-dbv1.syldb *.fastq -t (threads) > profiling.tsv
conda install -c bioconda sylph
Warning
conda install may break if AVX2 instructions are not available on your CPU. See the issue here. The binary and source install still work.
Requirements:
- rust (version > 1.63) programming language and associated tools such as cargo are required and assumed to be in PATH.
- A c compiler (e.g. GCC)
- make
- cmake
Building takes a few minutes (depending on # of cores).
git clone https://github.com/bluenote-1577/sylph
cd sylph
# If default rust install directory is ~/.cargo
cargo install --path . --root ~/.cargo
sylph query test_files/*
If you're on an x86-64 system, you can download the binary and use it without any installation.
wget https://github.com/bluenote-1577/sylph/releases/download/latest/sylph
chmod +x sylph
./sylph -h
Note: the binary is compiled with a different set of libraries (musl instead of glibc), probably impacting performance.
# all fasta -> one *.syldb; fasta are assumed to be genomes
sylph sketch genome1.fa genome2.fa -o database
#EQUIVALENT: sylph sketch -g genome1.fa genome2.fa -o database
# multi-sample sketching of paired reads
sylph sketch -1 A_1.fq B_1.fq -2 A_2.fq B_2.fq -d read_sketch_folder
# multi-sample sketching for single end reads, fastq are assumed to be reads
sylph sketch reads.fq
#EQUIVALENT: sylph sketch -r reads.fq
# ANI querying
sylph query database.syldb read_sketch_folder/*.sylsp -t (threads) > ani_queries.tsv
# taxonomic profiling
sylph profile database.syldb read_sketch_folder/*.sylsp -t (threads) > profiling.tsv
The pre-built databases available here can be downloaded and used with sylph for profiling and containment querying.
For common use cases and fast explanations, see the above cookbook.
For incorporating taxonomy and manipulating output formats, see the sylph-utils repository.
- Made unknown estimation (-u) more robust for low-depth short-read sequencing.
See the CHANGELOG for complete details.
Jim Shaw and Yun William Yu. Metagenome profiling and containment estimation through abundance-corrected k-mer sketching with sylph (2023). bioRxiv. (Accepted for publication)