ACR (Additional Clustering Refiner)

Additional Clustering Refiner (ACR), which regroups the contigs of the metagenome-assembled genomes (MAGs) using k-means clustering algorithm, to obtain MAGs with better quality.

Usage

Usage: acr.py -g [bin folder] -c [coverage file] -o [output]

Options:

  -h, --help            show this help message and exit
  -g GENOME, --genome=GENOME
                        genome file path
  -o OUTPUT, --output=OUTPUT
                        output folder
  -c COVERAGE, --coverage=COVERAGE
                        contig coverage file
  -e EXTENSION, --extention=EXTENSION
                        genome file extension |default: fa
  -p PREFIX, --prefix=PREFIX
                        prefix |default: refine
  -s SIZE, --size=SIZE  MAG size filter |default: 500k
  -t PROCESS, --process=PROCESS
                        number of workers | default = 1
  -b BYPASS, --bypass=BYPASS
                        bypass prodigal - hmmsearch | default = N
  -j JGI, --from_jgi_cov=JGI
                        please insert Y or N (Y=using jgi coverage file) |
                        default = N
  -m GMESEUK, --run_gmesEuk=GMESEUK
                        gene prediction with gmes for Euk (Y or N) | default =
                        Y
  --comp=SCORE_COMPLETENESS
                        core gene completeness | default = 50
  --cont=SCORE_CONTAMINATION
                        core gene contamination | default = 10
  --target=TARGET       refiner target (Prok or Euk) | default = Both
                        (Prokaryote and Eukaryote)

Examples

python acr.py -g test_SHIPPO/bin/ -o test_SHIPPO/refine -c test_SHIPPO/cov.txt

The example data are genome bins from Seong, et. al's previous ocean microbiome study.

Requirements

Please download core gene database file to ACR path. Please check thecommands below.

#in the [ACR path]
wget -O data.tar.gz https://figshare.com/ndownloader/files/41282157
tar -zxvf data.tar.gz

Check the Python library

ACR requires python 3.7 or above.

Here are the python library requirements

eukcc==0.2
EukRep==0.6.7
scikit-learn==0.19.2
pandas==1.1.2
kmeans1d==0.3.1

1. conda (or use mamba) create -n ACR -c bioconda python=3.7 eukcc==0.2 EukRep==0.6.7 scikit-learn==0.19.2 pandas==1.1.2
2. conda activate ACR
3. pip install kmeans1d

To run ACR, the absolute paths of prodigal and hmmsearch must be written in the program.txt file as follows:

e.g)
prodigal:[/usr/bin/prodigal]
hmmsearch:[/usr/bin/hmmsearch]
hmmalign:[add absolute path]
hmmpress:[add absolute path]
pplacer:[add absolute path] #https://matsen.fhcrc.org/pplacer/
guppy:[add absolute path] #https://matsen.fhcrc.org/pplacer/
EukRep:[add absolute path] #https://github.com/patrickwest/EukRep
runGMES:[add absolute path] #please install GeneMark-ES [http://exon.gatech.edu/GeneMark/license_download.cgi]

The overall schematic of the ACR refinement approach with detailed steps

-Overall ACR workflow

Benchmark with CAMI II rhizosphere data

A) MQ and HQ MAGs of different genome binner intersections using CAMI2’s rhizosphere gold-standard assembly dataset.

B) Recovered fungal MAGs’ genome quality statistics and MAGs information via ACR.

hoonjeseong / acr