Protein Feature Extraction for Machine Learning

Python code to extract features from Protein sequences for Machine Learning/Deep Learning

Protein feature extraction is carried out using Biopython package

Format:

Features (27 features):

1. AA-count (20x features)
2. aromaticity (1x)
3. secondary_structure_fraction (3x)
4. isoelectric_point (1x)
5. molecular_weight (1x)
6. instability_index (1x)

Packages required (other than built-in) for the execution of code... -Pandas -pickle -Biopython -subprocess

Top N features for identifying Insuliin protein sequence

Format:

Installation

For windows Windows users have to specify the path to fasta files and output folder in linux style of referencing directory using / slash rather than \ eg C:/folder_name/file_name.fasta This issue will be fixed in future updates

pip install discere

For linux

pip3 install discere

Usage

  import discere.discere as di
  
  di.extract_feature('./Documents/positive_training.fasta', 
                     './Documents/negative_training.fasta', 
                     './Documents')

di.extract_feature(input_file1, input_file2, output_directory)

output

Outputs are stored in user_specified_path/output in .txt, .arff and .csv formats