Jasmine969 / qshgm-code

Outline for README

• Package Overview
• Run Scripts
• Requirements

Package Overview

├─data
│  ├─source
│  │  ├─aac
│  │  │      negative_aac.csv
│  │  │      positive_aac.csv
│  │  └─fasta
│  │         negative.fasta
│  │         positive.fasta
│  └─train
│          data.csv
│          label.csv
│          merge.csv
│  model.py
│  nn.py
│  out.txt

• model.py
  • the script used for training samples with SVM, KNN and RF (random forest)
• nn.py
  • the script used for training samples with Neural Network
• data/source/fasta
  • raw FASTA file
• data/source/aac
  • the Amino Acid Composition (AAC) calculates the frequency of each amino acid type in a protein or peptide sequence
• data/train
  • data.csv -- merge positive_aac.csv and negative_aac.csv, used for the script model.py
  • label.csv -- mark the corresponding positive and negative samples, used for the script model.py
  • merge.csv -- merge data.csv and label.csv, used for the script nn.py
• out.txt
  • the results we got

Run Scripts

model.py

python model.py -e ESTIMATOR [-dp DATA_PATH] [-lp LABEL_PATH]

• -e, required, the estimator will be used, include ['svm', 'knn', 'rf']
• -dp, the path of data.csv
• -lp, the path of label.csv

nn.py

python nn.py [-e EPOCHS] [-bs BATCH_SIZE] [-lr LEARNING_RATE] [-dp DATA_PATH]

• -e, epochs, default=30
• -bs, batch_size, default=64
• -lr, learning_rate, default=0.01
• -dp, the path of merge.csv

Requirements

In the experiment, all scripts run in python 3.7, and the third-party packages we used are listed below:

numpy		1.16.2
pandas		0.25.0
sklearn		0.23.2
torch		1.4.0+cpu
torchvision	0.5.0+cpu