Name: WIsH
Owner: Söding Lab
Description: Predict prokaryotic host for phage metagenomic sequences
Created: 2017-02-14 16:45:46.0
Updated: 2017-10-10 13:20:28.0
Pushed: 2018-01-15 17:58:59.0
Homepage: null
Size: 1677
Language: C++
GitHub Committers
| User | Most Recent Commit | # Commits |
|---|
Other Committers
| User | Most Recent Commit | # Commits |
|---|
WIsH is licensed under the General Public License (see the LICENSE file). Copyright Clovis Galiez (clovis.galiez@mpibpc.mpg.de).
Build the docker container:
sh
path/to/repository/of/WiSH
er build -t wish .
To run WIsH from the container:
er run -v /some/host/folder:/data wish <some WiSH commands>
sh
clone https://github.com/soedinglab/WIsH.git
IsH
e .
Compiling with OpenMP support on MacOS requires a recent gcc compiler. You can get gcc from homebrew.
sh
install gcc@6
rt CC=gcc-6
rt CXX=g++-6
clone https://github.com/soedinglab/WIsH.git
IsH
e .
If you want to enjoy the parallelization of WIsH, you should have the OpenMP library installed. WIsH uses C++11. The model construction and the interaction prediction are both parallelized. In both cases, it spreads one bacterial genome/model per thread as soon as you give WIsH the number N of threads to use (with the parameter “-t N”).
You need two different directory containing only sequence data in FASTA format. One should contain your potential host genomes (one becteria per FATSA file), the other should contain the viral contigs/genomes (one virus per FATSA file).
To run a prediction, you should proceed in two steps:
1 - Create the models from the bacterial genomes you stored in FASTA format in prokaryoteGenomesDir:
sh
r modelDir
sH -c build -g prokaryoteGenomesDir -m modelDir
This will create a model in modelDir for every bectrial genome.
2 - Run the prediction on the viral sequences you stored in FASTA format in phageContigsDir:
sh
r outputResultDir
sH -c predict -g phageContigsDir -m modelDir -r outputResultDir -b
This will output a file llikelihood.matrix containing a matrix of log-likelihood (rows are a bacteria, and columns are a viral contigs), and a “summary” file prediction.list containing for every viral sequence the host corresponding to highest log-likelihood (-b option).
The files can be further analyzed with any text editor or with R:
read.table("outputResultDir/llikelihood.matrix")
ictions = read.table("outputResultDir/prediction.list")
ow the number of viral contigs targeting every potential hosts:
e(predictions$V2)
ow the histogram of the log-likelihoods for the best predictions:
(predictions$V3)
If you want to get p-values for your predictions, you need to know the null parameters for a new bacterial model. To get them, you must run the predictions on a large set of phage genomes that are know not to infect your bacterial model (let's call it the null set of phages) and use the prediction likelihood to fit the null-model parameters. You can use the script computeNullParameters.R that takes the predictions on this null set of phage and create a file containing the null parameters for every bacterial model. To get the p-values while predicting interactions, please specify the options “-b -n nullParameters.tsv” in a prediction call.