search

santirdnd / COPLA

A taxonomic classifier of plasmids
https://castillo.dicom.unican.es/copla/
GNU General Public License v3.0
25 stars 4 forks source link

readme

COPLA, a taxonomic classifier of plasmids

How to install COPLA

Dependencies

COPLA is a pipeline of scripts programmed with bash, python, perl, and ruby. In addition to the respective runtimes, COPLA requires several third-party programs and Python libraries to function. The following list details the specific versions used during the development of COPLA:

Installation

First we need to download the scripts and databases used by COPLA from GitHub:

# Change PROJECT_ROOT_DIRECTORY value to suit your preferences
PROJECT_ROOT_DIRECTORY=copla
git clone https://github.com/santirdnd/COPLA ${PROJECT_ROOT_DIRECTORY}
cd ${PROJECT_ROOT_DIRECTORY}

A conda environment has been provided as a convenient way to install most of the COPLA dependencies. Just run the following command:

conda env create -f copla.environment.yml -n copla

Alternatively, run this other command if you want to use the exact environment used during COPLA development:

conda create --file copla.spec-file.txt -n copla

MacSyFinder, the program used for MPF typing, is a bit trickier to install as it is a Python2 program and multiple versions of Python can't be mixed in the same conda environment. A specific conda environment is used to facilitate its installation. Run the following commands to install MacSyFinder:

conda env create -f macsyfinder.environment.yml -n macsyfinder
conda activate macsyfinder
cd ${CONDA_PREFIX}
wget https://github.com/gem-pasteur/macsyfinder/archive/macsyfinder-1.0.5.tar.gz
tar zxf macsyfinder-1.0.5.tar.gz
wget https://github.com/gem-pasteur/macsyview/archive/macsyview-1.0.1.tar.gz
tar zxf macsyview-1.0.1.tar.gz
mv macsyfinder-macsyfinder-1.0.5 src
mv macsyview-macsyview-1.0.1 src/macsyview
cd src
python setup.py build
python setup.py test -vv
python setup.py install --prefix=${CONDA_PREFIX}
conda deactivate
cd ${PROJECT_ROOT_DIRECTORY}

Finally, we need to download the software for ANI calculation. We will use ani.rb, a Ruby script provided in the Enveomics Collection:

wget https://github.com/lmrodriguezr/enveomics/raw/master/Scripts/ani.rb
chmod +x ani.rb
mv ani.rb bin/

Once we have all the necessary software, we need to download the databases that COPLA uses. For this reason, the download_Copla_databases.sh script has been provided, which additionally generates the CoplaDB.fofn file in the databases/Copla_RS84 directory. At the end of the command execution, this file should contain a list with the absolute paths to the Fasta sequences that make up the database. Just be sure to adapt the contents of the copla.ini configuration file to your local installation so that CoplaDB.fofn is generated automatically:

bin/download_Copla_databases.sh
head databases/Copla_RS84/CoplaDB.fofn

A last script, post_install_test.sh, has been developed to verify the correct installation and operation of COPLA. If warning or error messages appear for any of the tests performed, it will be an indication that parts of the COPLA installation should be checked. Anyway, a success message in the last two tests would indicate that COPLA may well be installed correctly.

bin/post_install_test.sh

Other relevant infomation:

Using COPLA

COPLA can be used to predict the PTU to which a plasmid belongs. For this just its nucleotide sequence is needed:

bin/copla.py test/NZ_CP028167.1.fna \
    databases/Copla_RS84/RS84f_sHSBM.pickle \
    databases/Copla_RS84/CoplaDB.fofn \
    test/NZ_CP028167.1_output

Draft plasmids are supported by COPLA automatically executing a simple concatenation of their contigs.

If the plasmid ORFeome is uploaded by the user, it will be used by COPLA (option -a) to obtain a type MOB and MPF according to the user specifications.

Indicating the query topology (option -t, linear or circular) is recommended for more accurate MPF typing. If not provided, COPLA uses a circular topology by default. Multifasta input sequences are always processed using a linear topology. For further details see Abby et al. PMID: 26979785.

Finally, if the user provides the taxonomic data asociated with the plasmid host (options -k, -p, -c, -o, -f, -g, and -s), COPLA will warn the user about changes in the host range of the query's assigned PTU.

The following command shows a complete example of the use of COPLA:

bin/copla.py test/NZ_CP028329.1.fna \
    databases/Copla_RS84/RS84f_sHSBM.pickle \
    databases/Copla_RS84/CoplaDB.fofn \
    test/NZ_CP028329.1.fna_output \
    -a test/NZ_CP028329.1.faa \
    -t circular \
    -k Bacteria \
    -p Firmicutes \
    -c Bacilli \
    -o Lactobacillales \
    -f Lactobacillaceae \
    -g Lactobacillus \
    -s 'Lactobacillus sp. D1501'

How COPLA works

Just as sHSBM, COPLA infers the PTU membership from the similarity relationships between plasmids in the database. After calculating the ANI percentage identity between the query and the reference plasmid set, COPLA inserts the query into the reference network and performs an statistical search for similar plasmids. Finally, the query is assigned to a known PTU, or to a new PTU (labeled as PTU-?) if the algorithm find clues pointing to that outcome. PTUs will not be named for clusters with fewer than 4 members. For the user to evaluate the COPLA PTU assignation of a query plasmid, a score is provided based on the overlap of the graph partitions before and after the query was inserted into the reference network (for additional details see troubleshooting bellow).

COPLA output

COPLA output consists of five different files:

Troubleshooting: Predicted outcome types

COPLA predictions can result in three different outcomes for PTU membership of the query plasmid. A query could be a member of: (i) a known PTU, (ii) a putative new PTU (provisionally named as "PTU-?"), or (iii) the plasmid remains unclassified (displayed as "-"). The quality of these predictions can be further evaluated with the help of the score output. To assess COPLA performance, we sampled 1,000 RefSeq200 plasmids not present in the reference database (RefSeq84). As shown in Figure 1, results indicate that ~88% of all queries (spanning all three classes) the prediction score a >99%, with the remaining predictions scoring lower values as shown in the figure.

Score distribution

Figure 1. Score distribution for 1,000 plasmids sampled from RefSeq200, not present in the COPLA reference database (RefSeq84). The figure displays a semilogarithmic plot of the number of plasmids containing each given score.

To validate the PTU assignment, a score of 90% is recommended. This threshold means that, for a 10-member PTU, the AI has conficting data for clustering 1 of them. As an example, the problematic plasmid could be a cointegrate of plasmids belonging to two different PTUs. Assuming the 90% score threshold, COPLA confidently assigned 93% of the 1,000 samples.

To help understand the different results that COPLA provides, Figure 2 shows several representative schematics.

Representative prediction outcomes

Figure 2. Representative prediction outcomes. The query plasmid is represented by the node with the red inner circle. For all other nodes, the color of the inner circle represents the PTU assigned in the reference database (i.e. using only RefSeq84 plasmids). The outer rings represent the PTU assigned by COPLA. Yellow represent the PTU assigned to the query, green is a different PTU, and grey colors represent not assigned PTUs.

When the query is part of a small graph component (4 members or less) the PTU prediction is pretty much trivial (cases 1 to 4):

In cases 5 to 7 the query is part of a larger graph component (>4 members), that is, a known PTU. In these cases, the assignation of the query is straightforward. These cases represent the predominant scenarios in which no significant differences in graph partitioning were detected.

Cases 8 and 9 represent scenarios where the inclusion of the query leads to a different partition in the connected component to which the query belongs. Case 8 highlights the case of plasmids (such as cointegrate plasmids) for which the algorithm can not decide unambigously to which PTU should assign the query. Case 9 showcases situations of loosely defined PTUs (showing low intragroup density, such as PTU-FE see ref. [1]). The score provided by COPLA is key to detect and to assess the level of confidence of the PTU prediction.

References

If you find COPLA useful, please cite the following papers: