POEM is a pipeline which can predict operons and core operons from metagenomic genome/assembly or short reads. It can be run on most * NIX systems and Windows WSL.
This pipeline is available on Linux systems and Windows WSL. Make sure that you have the following installed on linux
-
Anaconda for Python 3.7
-
make sure to add path of conda to $PATH environment variable
Installation is simple once Anaconda and Conda are installed. Type or paste the following commands into your terminal in whichever subfolder you want to keep POEM.
$ git clone https://github.com/Rinoahu/POEM_py3k
$ cd ./POEM_py3k
$ bash ./install.sh
The installation script calls conda to install all the necessary python packages and software, as well as the COG database. To avoid conflict, it will create a new conda environment named POEM_py3k. PS: POEM calls either Prodigal or MetaGeneMark for gene predictions. If the users want to use MetaGeneMark, they must install it by themselves, because MetaGeneMark requires academic users to agree to an license before downloading. After installtation, make sure to add path of binary executable file gmhmmp in MetaGeneMark to $PATH environment variable.
- Install Ubuntu (or preferred * Nix distribution). Ubuntu can be downloaded for free from the windows app store
- Open Windows Powershell as administrator and enter the following
Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Windows-Subsystem-Linux
more information on windows subsystem linux can be found at https://docs.microsoft.com/en-us/windows/wsl/install-win10. Reboot if prompted 3. Run and initialize your Linux distribution (set password etc). Update it by entering the following if using Ubuntu, or your distribution's update commands.
sudo apt update && sudo apt upgrade
- Download Anaconda 3.7 for Linux (not windows). Run the installer in your Linux distribution (/mnt/c/...).
- Open a new terminal, create a subfolder wherever you want to store POEM, and enter the following commands
$ git clone https://github.com/Rinoahu/POEM_py3k
$ cd ./POEM_py3k
$ bash ./install.sh
The installation script calls conda to install all the necessary python packages and software, as well as the COG database. PS: POEM calls either Prodigal or MetaGeneMark for gene predictions. If the users want to use MetaGeneMark, they must install it by themselves, because MetaGeneMark requires academic users to agree to an license before downloading. After installtation, make sure to add path of binary executable file gmhmmp in MetaGeneMark to $PATH environment variable.
note: Windows WSL will not allow linux programs with GUIs to run, and the file systems are separate. Windows programs will not be able to access files in your Linux system unless you move them back to your Windows files. To connect to your existing drive use /mnt/c/User/.... if you are using Cytoscape to visualize your networks, install Cytoscape for windows not Linux. To access the network files you will need to move them to your windows drive.
The example directory contains two genome fasta files, run runme.sh to test the pipeline.
To check if POEM is installed correctly, the users can make a quick and simple test:
$ cd ./example
$ bash ./runme.sh ecoli.fasta.bz2
This should output a network with 5 pairs of genes if you check the .sif file in Cytoscape. note: this is to verify your installation is working correctly only and is not representative of POEMs intended function (this is looking at a single genome rather than metagenomic data, and is identifying duplicated genes). A short demo using actual metagenomic data is included along with a guide in the demo folder
To evaluate the performance of POEM on real data, the users can test it on the file of full.fasta.bz2:
$ cd ./example
$ bash ./runme.sh full.fasta.bz2
The full.fasta.bz2 file contains 48 whole genomes of bacteria, and it will take several hours.
POEM is recommended for finding operons in preassembled metagenomic data, but can also accept short or long reads, for which it uses IDBA_UD to assemble
For preassembled fasta files
$ bash ./bin/run_poem.sh -f file.name -a n -p pro
or to use metagenemark for annotation instead of prodigal (if installed)
$ bash ./bin/run_poem.sh -f file.name -a n -p gmk
If you want to upload files to genbank you can also call Prokka to generate genbank files (this can extend runtime significantly)
$ bash ./bin/run_poem.sh -f file.name -a n -p pka
For short reads <600bp:
$ bash ./bin/run_poem.sh -f file.name -a y -p pro -l n
file.name is single fasta file. If the reads are paired-end files in fastq or fasta format, use the fq2fa command of IDBA_UD to convert them to a single fasta file. Interleaved paired-end reads in a single file are ok
For long reads >600bp:
$ bash ./bin/run_poem.sh -f file.name -a y -p pro -l y
file.name is single fasta file. If the reads are paired-end files in fastq or fasta format, use the fq2fa command of IDBA_UD to convert them to a single fasta file. Interleaved paired-end reads in a single file are ok
-f: Specifies input file name. File should be a single fasta format file, but file name is not important.
-a: Assembly mode. For preassembled files -a n/N, for reads (single end or interleaved paired ends) -a y/Y
-p: Gene prediction method. For Prodigal -p pro/prodigal, for metagenemark (if installed) -p gmk/genemark. To generate genbank files with prokka -p pka/prokka
-l: Read length: If assembly mode is on IDBA_UD will default to short read mode, which can also be specified by -l n. for long reads (>600bp) -l y
POEM will create a directory named filename.fasta_output to save the results. The results include serveral files:
1. input.fsa:
The contigs or scaffolds of IDBA-UD output in fasta format
2. input.fsa_gmk_aa.fsa:
The fasta file of protein sequence and gff file generated by prediction of MetaGeneMark on input.fsa from step 1.
Note: If Metagenemark is used to predict the gene, there will be an additional file named input.fsa.gff in the directory.
3. input.fsa.cdhit and input.fsa.cdhit.clstr:
The cd-hit clustering results of input.fsa_gmk_aa.fsa from step 2
4. input.fsa_aa.fsa:
Filtered input.fsa_gmk_aa.fasta according to the clustering result of step 3
5. input.fsa.flt.blast8:
Blastp results in -m 8 tabular format of input.fsa_aa.fsa from step 4 against COG database
6. input.fsa.flt.blast8.flt:
Filtered blastp output of step 5. The hits with identity < 80% are removed
7. input.fsa.cog:
COG annotations for the protein sequences of step 4. The first row of this tab-delimited file is the gene number. The rest of the file contains 3 columns: col 1 is the gene identifier, chromosome|scaffold name and cluster id split by "$$", col 2 is the COG identifier and col 3 is the functional category of the genes, for example:
uniq gene 3679
gene cog annot
CPIBHFJP_04443|Prodigal:2.6|3106475_aa|-|4659209|4659544$$gi|985000614|gb|CP014225.1|$$>Cluster_3671$$* COG1455 :G::Phosphotransferase system cellobiose-specific component IIC::Carbohydrate transport and metabolism
CPIBHFJP_04442|Prodigal:2.6|3106198_aa|-|4658880|4659158$$gi|985000614|gb|CP014225.1|$$>Cluster_3846$$* COG1455 G::Phosphotransferase system cellobiose-specific component IIC::Carbohydrate transport and metabolism
CPIBHFJP_04441|Prodigal:2.6|3105940_aa|-|4658384|4658734$$gi|985000614|gb|CP014225.1|$$Orphan$$Orphan COG1447 G::Phosphotransferase system cellobiose-specific component IIA::Carbohydrate transport and metabolism
...
8. input.fsa.locus:
This file records the gene's locus on genome. This file contains 5 columns: col 1 is gene identifier and chromosome|scaffold name split by "$$", col2 is the chromosome|scaffold name, col3 is the strand of the gene, col 4 and 5 are the start and end of the gene. For example:
CPIBHFJP_00001|Prodigal:2.6|772_aa|+|255|857$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 255 857
CPIBHFJP_00002|Prodigal:2.6|1014_aa|+|883|1308$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 883 1308
CPIBHFJP_00003|Prodigal:2.6|1165_aa|+|1586|1693$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 1586 1693
...
9. input.fsa.adjacency:
This file is the predicted operonic adjacent genes. This file contains 11 columns: col 1 is the gene 1's id; col 2 is the chromosome|scaffold id where gene 1 is located; col 3-5 are strand, start and end of gene 1; col 6 is gene 2's id; col 7 is the chromsome|scaffold id where gene 2 is located; col 8-11 are strand, start and end of gene 2. For example:
CPIBHFJP_00001|Prodigal:2.6|772_aa|+|255|857$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 255 857 CPIBHFJP_00002|Prodigal:2.6|1014_aa|+|883|1308$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 883 1308 False
CPIBHFJP_00002|Prodigal:2.6|1014_aa|+|883|1308$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 883 1308 CPIBHFJP_00003|Prodigal:2.6|1165_aa|+|1586|1693$$gi|985000614|gb|CP014225.1| gi|985000614|gb|CP014225.1| + 1586 1693 False
...
10. input.fsa.operon:
The predicted full operons. This file contains 3 columns: col 1 is the ids of operonic genes split by --> or <--, arrow stands for the strand of gene; col 2 is the COG ids of the operonic genes; col 3 is the COG annotation of the operonics genes split by --> or <--. For example:
gene_id COG_id annotation
CPIBHFJP_00007|Prodigal:2.6|2673_aa|+|3821|3946$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00008|Prodigal:2.6|2862_aa|+|3961|4182$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00009|Prodigal:2.6|3078_aa|+|4218|4484$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00010|Prodigal:2.6|3373_aa|+|4484|4867$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00011|Prodigal:2.6|3449_aa|+|4887|5078$$gi|985000614|gb|CP014225.1| unknown-->unknown-->unknown-->COG3001-->COG3001 unknown::unknown-->unknown::unknown-->unknown::unknown-->G::Fructosamine-3-kinase::Carbohydrate transport and metabolism-->G::Fructosamine-3-kinase::Carbohydrate transport and metabolism
CPIBHFJP_00036|Prodigal:2.6|20592_aa|+|29823|30533$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00037|Prodigal:2.6|21036_aa|+|30538|31215$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00038|Prodigal:2.6|21527_aa|+|31230|31937$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00039|Prodigal:2.6|21840_aa|+|31937|32485$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00040|Prodigal:2.6|22830_aa|+|32495|33661$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00041|Prodigal:2.6|23958_aa|+|33634|35169$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00042|Prodigal:2.6|24099_aa|+|35169|35822$$gi|985000614|gb|CP014225.1|-->CPIBHFJP_00043|Prodigal:2.6|25233_aa|+|35889|37196$$gi|985000614|gb|CP014225.1| COG0398-->unknown-->COG0398-->COG2128-->COG4134-->COG4135-->COG4136-->COG2897 S::Uncharacterized membrane protein YdjX, TVP38/TMEM64 family, SNARE-associated domain::Function unknown-->unknown::unknown-->S::Uncharacterized membrane protein YdjX, TVP38/TMEM64 family, SNARE-associated domain::Function unknown-->P::Alkylhydroperoxidase family enzyme, contains CxxC motif::Inorganic ion transport and metabolism-->R::ABC-type uncharacterized transport system YnjBCD, periplasmic component::General function prediction only-->R::ABC-type uncharacterized transport system YnjBCD, permease component::General function prediction only-->R::ABC-type uncharacterized transport system YnjBCD, ATPase component::General function prediction only-->P::3-mercaptopyruvate sulfurtransferase SseA, contains two rhodanese domains::Inorganic ion transport and metabolism
11. input.fsa.cog_adjacency:
This file contain two parts: The first part converts operonic adjacency to COG id adjacency and statistic of the COG id adjacency. The second part evaluate the core operon and its closest real operon.
##############################
# COG adjacency
##############################
geneA geneB cA cB cAB cA+cB-cAB cAB/(cA+cB-cAB)
COG3188 COG3539 11 32 7 36 0.194444444444
COG2801 COG2963 16 16 11 21 0.52380952381
COG1662 COG3677 10 9 8 11 0.727272727273
...
##############################
# evalute the core operon
##############################
predict_operon real_operon_name real_operon_cog Precise Recall F1 predict_annot real_annot
COG0601$$COG1173 sso:SSO3058$$sso:SSO3059 COG0601$$COG1173 1.000000 1.000000 1.000000 ABC-type dipeptide/oligopeptide/nickel transport system, permease component$$ABC-type dipeptide/oligopeptide/nickel transport system, permease component ABC-type dipeptide/oligopeptide/nickel transport system, permease component$$ABC-type dipeptide/oligopeptide/nickel transport system, permease component
COG1129$$COG1172 eco:b1513$$eco:b1514$$eco:b1515$$eco:b1516$$eco:b1517$$eco:b1518$$eco:b1519 COG1129$$COG1172$$COG1172$$COG1879$$COG1830$$COG1359$$COG4106 1.000000 0.428571 0.600000 ABC-type sugar transport system, ATPase component$$Ribose/xylose/arabinose/galactoside ABC-type transport system, permease component ABC-type sugar transport system, ATPase component$$Ribose/xylose/arabinose/galactoside ABC-type transport system, permease component$$Ribose/xylose/arabinose/galactoside ABC-type transport system, permease component$$ABC-type sugar transport system, periplasmic component, contains N-terminal xre family HTH domain$$Fructose-bisphosphate aldolase class Ia, DhnA family$$Quinol monooxygenase YgiN$$Trans-aconitate methyltransferase
...
12. input.fsa.core_cog_adjacency:
The core COG adjacency is the same as the first part of file from step 11. This file is tab delimited.
geneA geneB cA cB cAB cA+cB-cAB cAB/(cA+cB-cAB)
COG1173 COG0601 5 5 5 5 1.0
COG1172 COG1129 8 8 6 10 0.6
COG2963 COG2801 16 16 11 21 0.52380952381
COG1662 COG3677 10 9 8 11 0.727272727273
13. input.fsa.core_network.sif, input.fsa.core_node.tab:
Network, node attribute and edge attribute extracted from step 11 for cytoscape visualization. Figure 1 shows an example to view the core operons by cytoscape.
To visualize the network in Cytoscape import the network from the .sif file. To add the gene prediction annotations, import the input.fsa.core_node.tab file as a table. Change the node labels to function/passthrough mapping under style. Optionally change fill color to 'color' to colorcode by function, and change edge width to 'interaction' to visualize frequency of cooccurance of genes.
Figure 1: visualization of core operons in cytoscape