/Prediction-of-Anti-fungal-Peptides

Predicting Anti-fungal peptides by training a model using SVM Scikit Learn

Primary LanguagePython

IQB ASSIGNMENT 2

  • Team Members:-

    • Amandeep Kaur (2018014)
    • Deepi Garg (2018389)
    • Ria Gupta (2018405)

  • Python script: iqb.py

  • Input: 2 CSV format file (Default: test.csv, train.csv)

  • External support: GPS Raghava’s Pfeature (Pass train.csv(test.csv) to Pfeature to get dipeptide_result_train.csv(dipeptide_result_test.csv))

  • Output: 1 CSV format files (Default: output.csv)

  • Command to run the script:-

    Put train.csv, test.csv, dipeptide_result_train.csv, dipeptide_result_test.csv in the same folder as the code file iqb.py and in terminal use the following command.

    $ python3 iqb.py

    The python script uses the library sklearn to import svm.

  • We define a function dipeptite_composition(file), the function reads the file line-by-line and stores values of dipeptide compositions in a list of lists and returns it.

  • We define amino_composition(file, i), file is the input file on which function works/reads and i is the number of the column to be read as the column containing the amino acid sequence differs in the files, train.csv and test.csv. The function reads the amino acids sequences one-by-one, and for each sequence, creates a list which maintains the percentage frequency count of each amino acid in the sequence.

    Percentage frequency count (A) = (Frequency count(A)/Length of sequence)*100

    The function returns a list of such lists for each sequence.

  • TRAINING The labels given in train.csv are copied into a list Y. The methods amino_composition and dipeptide_composition are called on train.csv and values are stored in lists X1 and X2 respectively. For each sequence, the values in X1 are appended to X2. Now a SVM is trained with values X2 and Y with 1 / (n_features * X.var()) as the value of gamma. Once the SVM is trained, we proceed to test it.

  • TESTING The methods amino_composition and dipeptide_composition are called on test.csv and values are stored in lists x1 and x2 respectively. For each sequence, the values in x1 are appended to x2. Now we call the pre-defined function predict() on the trained SVM(named clf) and save results in the list ofile. These results are our output labels. We create a new file, output.csv and for each ID of amino acid sequence in test.csv we print the ID with its corresponding predicted label.