-
Notifications
You must be signed in to change notification settings - Fork 6
5‐minute sylph tutorial
After installation, clone this repository if you have not done so and run the following.
git clone https://github.com/bluenote-1577/sylph
cd sylph
## install sylph. see installation instructions
# cargo install --path . --root ~/.cargo
## OR
# conda install -c bioconda sylph
# sketch reads and genomes. fastq -> samples, fasta -> queries
sylph sketch test_files/o157_reads.fastq test_files/e.coli*.fa -o databaseThere are two types of files: *.syldb and *.sylsp.
-
FASTQ files are treated as samples and turn into
*.sylsp -
FASTA files are treated as genomes and turn into
*.syldb - For other options, such as paired-end reads, see the cookbook
Genomes are aggregated into one syldb file, and each genome is queried against all sylsps.
# query for ANI
sylph query o157_reads.fastq.sylsp database.syldb
# ALTERNATIVE: lazy containment without pre-sketching also works
# sylph query test_files/*The o157_reads.fastq is a "metagenomic sample" containing only E. coli O157 with 1x coverage and 95% identity reads (i.e. 5% error). We query the reads and the database consisting of multiple E. coli genomes.
You'll see something like the following
Sample_file Genome_file Adjusted_ANI Eff_cov ANI_5-95_percentile Eff_lambda Lambda_5-95_percentile Median_cov Mean_cov_geq1 Containment_ind Naive_ANI Contig_name
../test_files/o157_reads.fastq ../test_files/e.coli-o157.fasta 99.73 0.360 99.61-99.91 0.360 0.33-0.38 1 1.187 6208/21899 96.02 NZ_CP017438.1 Escherichia coli O157:H7 strain 2159 chromosome, complete genome
../test_files/o157_reads.fastq ../test_files/e.coli-EC590.fasta 98.25 0.319 98.06-98.55 0.319 0.29-0.34 1 1.172 3122/19330 94.29 NZ_CP016182.2 Escherichia coli strain EC590 chromosome, complete genome
../test_files/o157_reads.fastq ../test_files/e.coli-K12.fasta 98.16 0.327 97.96-98.47 0.327 0.29-0.35 1 1.171 3114/19485 94.26 NC_007779.1 Escherichia coli str. K-12 substr. W3110, complete sequence
These are statistics for each genome against our reads. The ANI can be interpreted as nearest-neighbour ANI searching, i.e., "what is the highest ANI between the genomes in my sample and the reference genome?". See here for a detailed output explanation.
- The 3rd column gives the coverage adjusted ANI between the genome and this sample.
- The second last column is the Naive ANI -- what you would approximately get without sylph's statistical adjustment (i.e. if you used Mash or Sourmash).
Notice the big difference between 1. and 2. This is because the reads are only 1x coverage: methods like mash screen and sourmash give biased ANI when coverage is low.
However, the Eff_cov gives smaller than 1x: this is because Eff_cov takes into account sequencing error. Sequencing error reduces the k-mer based coverages (sequencing errors invalidate k-mers). You can estimate the true coverage if you provide some more information. See the cookbook.
Here are the ANIs computed by skani between the three genomes:
test_files/e.coli-EC590.fasta 100.00 99.39 98.14
test_files/e.coli-K12.fasta 99.39 100.00 98.09
**test_files/e.coli-o157.fasta 98.14 98.09 100.00**So the ANIs should be 98.14, 98.09, and 100.0 for EC590, K12, and O157 respectively against the sample. As you can see, Sylph's estimates are quite good and much more reasonable than the Naive ANI.
In the above example, notice that querying each E. coli genome gave a high ANI value. However, only one E. coli genome is present in the sample, not all three.
Thus sylph query is not a profiler. It does not tell you the abundance of the genomes in your sample, just how similar your reference genome is to your metagenome.
To remove this redundancy, we can use the sylph profile instead.
> sylph profile test_files/* # or use 'profile' on the syldb and sylsp.
...
...
Sample_file Genome_file Taxonomic_abundance Sequence_abundance Adjusted_ANI Eff_cov ANI_5-95_percentile Eff_lambda Lambda_5-95_percentile Median_cov Mean_cov_geq1 Containment_ind Naive_ANI Contig_name
../test_files/o157_reads.fastq ../test_files/e.coli-o157.fasta 100.0000 100.0000 99.73 0.360 99.61-99.91 0.360 0.33-0.38 1 1.187 6208/21899 96.02 NZ_CP017438.1 Escherichia coli O157:H7 strain 2159 chromosome, complete genomeNotice the new Sequence_abundance and Taxonomic_abundance columns, which give the relative abundances as a percentage. See https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8184642/ for how taxonomic and sequence abundance differ.
There is only 1 genome in the sample, so it has 100% abundance.
Note
This is a simple example with multiple strains of a single species. For real metagenomes, we advise using databases that are dereplicated at the species level.
Our pre-built databases are all dereplicated at the species level.
See the tutorial here to learn how to profile against the GTDB database, an actual database of > 50,000 genomes prokaryotic genomes.