Home

Welcome to the Protein_Project wiki! Information about my dataset:-

The Dataset is of membrane alpha helices. The data file is in 3 lines per a protein sequence with a protein id, protein sequence and the feature sequence. The features have 3 possibilities Inside the cell, in the membrane or outside the membrane

Overview: Preparation:

Created a bash script and a git hub repository which has structure for my project

Main objects Extract the feature from your dataset Create cross-validated sets. Train a SVM using single sequence information, using sklearn Check different window sizes for the inputs Add evolutionary information by running psi-blast and extracting the information Train a SVM using multiple sequence information Optimize the performance of the SVM Analyze the results and compare it to previous work Use random forests and a simple decision tree and compare the performance with the SVM performance. Extract the data from 50 other proteins and test the performance Review the state of art for your predictor Write a report

In-depth steps:

1. Extract the feature from your dataset

Parsing: - A new python script was created and label parse.py. It objective is to open my data file at the beginning of the script and parse the file by sorting through the dataset text file membrane-alpha.3line.txt and separating the elements line by line. The separation of the different elements was done by the splitting the elements into their own list with the condition if/else of the remainder by the modulator of 3.

• id labels → idlist
• sequences → seqlist
• features → feat_list

The lists were indexed with the use of an enumeration of the list. The different elements were simulatanoesly written to a output file in the forloop – idlist.txt, seqlist.txt and feat_list.txt.

Note: I am to decide if to leave this as a script or if to define it as a function. This is a question I going to decide on as I code more as I want to make it easy to reuse code and also to create flow in the final program.

2.Create cross-validated sets.

Separating into train and test datasets:

To extract the features using the sci-kit learn OneHotencoder or the Dictvectorizer into a sparse matrix format on which I can use for an input into sklearn.

Either Using the cross_valid_sort in scikit learn I will need to divide the dataset into 3 or 5 different file on which to use for cross-validation in order to train and test a SVModel. Train a SVM on sequence info with sklearn. After which the model will be tested on a test set.

to be continued………...