This is the official code for our paper "PB-LKS: a python package for predicting Phage-Bacteria interaction through Local K-mer Strategy"
+--------------------------------------------+
| |---\ |----\ | | / /----\ |
| | | | | | | / | |
| |___/ |____/ -- | |/\ \----\ |
| | | \ | | \ | |
| | |_____/ |_____ | \ \____/ |
+--------------------------------------------+
- install conda(to manage environment)
- Change directory to the path of this project
cd {your_path_to_PBLKS} - Run following codes in your terminal
conda create -n PB-LKS python=3.10 conda deactivate (if base environment is activated) conda activate PB-LKS pip install -r requirements.txt
Both a command line tool and a python package is provided in this project. Developers and researchers may select according to needs after setting local environment.
A command line interface is implemented and can be used by running following commands in your terminal.
cd {your_path_to_PBLKS}
# display help message of PB-LKS CLI
python PB-LKS.py -h
# run and display result with example input
python PB-LKS.py -eother detailed arguments and their usages are listed as follows
usage: python PB-LKS.py [-h] [-p PHAGE] [-b BAC] [-e] [-xgb] [-fea] [-ba] [-o OUTPUT] [-d]
options:
-h, --help show this help message and exit
-p PHAGE, --phage PHAGE
path/folder to your phage sequence file(in fasta format)
-b BAC, --bac BAC path/folder to your bacteria sequence file(in fasta format)
-e, --example run model with example input and quit
-xgb, --xgboost run prediction with xgboost(default model is based on RandomForest)
-fea, --feature show 10 features of most importance and exit
-ba, --batch run prediction in batch, -p,-b should be folder if set True
-o OUTPUT, --output OUTPUT
the folder you want to save batch prediction results (results will be printed into terminal by default)
-d, --detail if set true, print detailed prediction results(single prediction), including important features and decision path.To make it more convenient for researchers and developers to use our model in more personalized tasks, a python package is also implemented. following functions are available
example code:
import PBLKS
bacteria_path = 'example1.fasta'
phage_path = 'example2.fasta'
# directly gives the result of prediction
result, prob = PBLKS.predict(bacteria_path, phage
7375
_path)
# gets descriptors from files
# can be used to reconstruct or train other type of models
feas = PBLKS.get_descriptor(bacteria_path, phage_path)
# predicts the interaction of each bacteria-phage pair in both folder
bac_folder = "your_folder_path"
phage_folder = "your_folder_path"
out_put_dir = "your_dest_folder"
# output'll be directed to dest folder, creating 2 result files
PBLKS.predict_in_batch(bac_folder, phage_folder, output_dir=out_put_dir)
# printing results into terminal
PBLKS.predict_in_batch(bac_folder, phage_folder, output_dir=None)
# get decision path
decision_path = PBLKS.show_decision_path(bacteria_path, phage_path)
# get top 10 important features
# result is orgnized in dict like: {kmer: importance}
features = PBLKS.show_feature_importance(top_cnt=10)Interpertability is a key feature in decision tree based alogrithms, several interfaces to visualize how our model make decisions are thus implemented.
Except calling -d or PBLKS.show_decision_path to show decision path with 01, we implemented a script to visualize decision trees in our model.
Users can run following command in termianl
python make_tree_plots.pyThis'll create a detailed visualization of every decision tree into /PBLKS/Trees, providing a more straightforward way to visualize decision.
A xgboost based model is also included in /models.
can be used by adding -xgb argument in CLI or adding parameter use_xgboost=True when calling methods in our package
Visit this git repository to get source code of PB-LKS:
https://github.com/wanchunnie/PB-LKS