fairy - fast approximate contig coverage for metagenomic binning

Fairy computes multi-sample contig coverage for metagenome-assembled genome (MAG) binning.

Fairy is used after metagenomic assembly and before binning. It can

Calculate coverage 100x-1000x faster than read alignment (e.g. BWA)
Give comparable bins for multi-sample binning (short read or nanopore reads)
Output formats that are compatible with MetaBAT2, MaxBin2, SemiBin2, and more

Caveats:

Don't use fairy for single-sample binning
Don't use fairy for PacBio HiFi

Note

See here for results and additional information/context about fairy.

Install (current version v0.5.5)

Option 1: conda install

mamba install -c bioconda fairy
# conda install -c bioconda fairy

Warning

If you're using linux, AVX512 instructions (e.g. a newer CPU) may be required for conda. Source install (option 2) and the static binary (option 3) will work for any CPU.

Option 2: Build from source

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/fairy
cd fairy

# If default rust install directory is ~/.cargo
cargo install --path . 
fairy -h

Option 3: Pre-built x86-64 linux statically compiled executable

If you're on an x86-64 Linux system, you can download the binary and use it without any installation.

wget https://github.com/bluenote-1577/fairy/releases/download/latest/fairy
chmod +x fairy
./fairy -h

Note: the binary is compiled with a different set of libraries (musl instead of glibc), probably impacting performance.

Quick start

Step 1: Index reads

# sketch/index short reads
fairy sketch -1 *_1.fastq.gz -2 *_2.fastq.gz -d sketch_dir

# sketch/index long reads
fairy sketch -r long_reads.fq -d sketch_dir

# rename the sketches if filenames are identical
fairy sketch -r dir1/reads.fq dir2/reads.fq -S sample1 sample2 -d sketch_dir

Step 2: Calculate coverage

# calculate coverage
fairy coverage sketch_dir/*.bcsp contigs1.fa -t 10 -o coverage1.tsv
fairy coverage sketch_dir/*.bcsp contigs2.fa -t 10 -o coverage2.tsv

Step 3: Bin

# default format is compatible with metabat2
metabat2 -i contigs1.fa -a coverage1.tsv ...
metabat2 -i contigs2.fa -a coverage2.tsv ...

# maxbin2 (requires different options; see below)
maxbin2 ...

# SemiBin2 (requires different options; see below)
SemiBin2 single_easy_bin -i contigs1.fa -a cov_aemb_1.tsv cov_aemb_2.tsv ...

Output

MetaBAT2 format (default)

The default output is compatible with the jgi_summarize_bam_contig_depths script from MetaBAT2 (the column names are different, however).

contigName  contigLen  totalAvgDepth  reads1.fq  reads1.fq-var  reads2.fq  reads2.fq-var  ...
contig_1    38370      1.4            1.4        1.1100          0       0
...

First three columns give the name, the length, and average coverage.
The next columns are mean coverage and coverage variance for each sample.

The above output can be fed directly into MetaBAT2 with default parameters.

SemiBin2 format (--aemb-format option)

Since fairy v0.5.5 and SemiBin v2.1, you can use SemiBin as follows

fairy coverage contigs1.fa reads1.bcsp --aemb-format -o cov_aemb1.tsv
fairy coverage contigs1.fa reads2.bcsp --aemb-format -o cov_aemb2.tsv
...
SemiBin2 single_easy_bin -i contigs.fa cov_aemb*.tsv -o results

Tip

Fairy usage for SemiBin2 is different than other tools: SemiBin2 requires separate coverage files for each read sample -- other tools require a single coverage matrix.

MaxBin2 format

Alternatively, --maxbin-format works directly with MaxBin2 and is also available. This removes the variance columns as well as the contigLen and totalAvgDepth columns.

Citing fairy

Jim Shaw, Yun William Yu. Fairy: fast approximate coverage for multi-sample metagenomic binning (2024). Microbiome.