ML algorithms and performance measures

1. The bash environment should be installed.

2. Phyton 2.7 executables should be accessible from the PATH environment variable.

3. Install scikit-learn, Pandas and scipy

1. export attCsynth_DIR=$YOUR_DIRECTORY/attCsynthPck

2. export attCsynth_SCRIPTS=$attCsynth_DIR/tools/scripts

1. Type the command below in your terminal and follow the instructions

python $attCsynth_SCRIPTS/Synthetic_attC_site_generator.py

2. The program will generate a fasta file in your current location

1. Type the command below in your terminal

cd $attCsynth_DIR/data/

python $attCsynth_SCRIPTS/RandomForestRegression.py -t L1_listABCD_input_file.undersampling -i attCFeatures.input \
-o attCFrequencies.output --v

2. To change default parameters of RandomForestRegression.py do:

   python $attCsynth_SCRIPTS/RandomForestRegression.py -help

1. We equilibrated our dataset by selecting an equal number of enriched and depleted mutants. To run undersampling procedure do:

cd $attCsynth_DIR/data/

python $attCsynth_SCRIPTS/underSampling.py -i L1_listABCD_input_file.txt -o L1_listABCD_input_file.undersampling

2. To execute underSampling.py for all feature lists do:

bash $attCsynth_SCRIPTS/sh/underSampling.sh input_file.txt

3. An undersampling file will created for each file list with extension .undersampling. you can also change default parameters
   

See how change default parameters of underSampling.py by doing:

python $attCsynth_SCRIPTS/underSampling.py -help

1. We tested four regression algorithms: Decision Tree Regression, Support Vector Regression, Ridge regression and Random forest regression, by applying three different feature selection strategies: PCA, k-best feature, and a manually feature selection (lists A to D, and their combinations)

   
   

    To run all algorithms type in your terminal

   
   

python $attCsynth_SCRIPTS/evaluateML.py -i L1_listABCD_input_file.undersampling -o L1_listABCD_output_file --v

2. The file L1_listABCD_output_file.reg will be created containing the evaluated measures: Pearson correlation coefficients, the mean absolute errors, the root mean square errors and the explained variance scores for each regression method.

3. To change default parameters of evaluateML.py do:

python $attCsynth_SCRIPTS/evaluateML.py -help