shishiyuan / ibpred Goto Github PK

Ion binding proteins (IBPs) and non-IBPs are stored in ./datas/data.fasta without labels. But first 114 sequences are IBPs, and other 207 samples are non-IBPs.

Create ibpred env, execute following commands in base env of Anaconda:

conda env create -n ibpred -f requirements.yaml
conda activate ibpred

Anaconda (Anaconda3-2021.11-Linux-x86_64) virtual environment called ibpred on Linux system was used.

To build ibpred, please execute following commands in base environment of Anaconda:

conda create -n ibpred -y
conda activate ibpred
conda install python==3.9.7 numpy==1.21.2 pandas==1.3.5 scipy==1.7.1 scikit-learn==1.0.1 tenacity==8.0.1 matplotlib==3.4.3 seaborn==0.11.1 ipykernel==6.4.1 traitlets==4.3.3 black -y
conda install -c conda-forge scikit-optimize==0.9.0 -y

If the user wants to run Jupyter notebook version-based workflow, all in process.ipynb, the ipykernel need to be installed.

The Jupyter notebook in VS code editor required; traitlets is optional for solving some bugs that can't run 2+ notebooks in the same env

However, failed :(

Otherwise you can reproduce the paper results by run the following command line in code directory (It's better to use ipynb because our results were produced by running all codes in ipynb):

python process.py

The matplotlib, seaborn, black packages are alternative for model training, but matplotlib and seaborn are necessary for plot pictures, as well black can be used for formatting python code.

auto_pred.py is the tool to identify IBPs. Users must confirm the model and the feature subset num used for prediction. Because the default parameter model is the base_bayes_clf_PD and default set_num is 193.

Command line in the ibpred env like this:

python auto_pred.py --file 'input_file.fasta' --model 'base_bayes_clf_PD' --set_num 193 --out 'result.csv'

Notice:

• Don't use the same identifiers for different samples.
• Don't place the same sequences in the input file.