/panclassif

A method to improve TCGA pancancer classifiers performance

Primary LanguagePythonMIT LicenseMIT

PanClassif: A machine learning classifier pipeline for TCGA pancancer classification

This is a complete machine learning pipeline package to work with TCGA cancer RNA-seq gene count data.

Data prerequisition

  • TCGA cancer & normal samples downloaded using TCGA2STAT
  • smoothed version of the above collected data using knn-smoothing (Wagner et al., 2017)

Functions

featSelect(homepath, cancerpath, normalpath, k)

Params

  • homepath : (str) Path where you want to save all the generated files and folders.
  • cancerpath : (str)
    Path where all the cancer's cancer gene expression matrix are located.
  • normalpath : (str)
    Path where all the cancer's normal gene expression matrix are located.
  • k : (int) The number of top genes you want to choose per cancer. (default: k=5) you can not put k less than 5

dataProcess(homepath,names,cancerpath,smoothed_cancer,smoothed_normal)

Params

  • homepath : (str) Path where you want to save all the generated files and folders.
  • cancerpath : (str) Path where all the cancer's cancer gene expression matrix are located.
  • names : (list) List of the cancer names found from featSelect function.
  • smoothed_cancer : (str) Path where all the cancer's smoothed cancer gene expression matrix are located.
  • smoothed_normal : (str) Path where all the cancer's smoothed normal gene expression matrix are located.

upsampled(names, homepath)

binary_merge(names, homepath)

multi_merge(names, homepath)

Params

  • names : (list) List of the cancer names found from featSelect function.
  • homepath : (str)
    Path where you want to save all the generated files and folders.

classification(homepath, classifier, mode, save_model)

Params

  • homepath : (str) Path where you want to save all the generated files and folders
  • classifer : (sklearn's classification model) Provide the classification model's instance you want to use. For example: RandomForestClassifier(n_estimators=100).
  • Or, classifer : (str) If you want to use "Neural Network" then just type "NN". For example: classifier = "NN"
  • mode : (str) There is two mode 1) binary 2) multi. Use "binary" for binary classification & "multi" for multiclass classification. (default: mode = "binary")
  • save_model : (str) Optional parameter. Use it only if you want to save the model. For example: save_model = "your_model_name"

gsea(homepath)

  • homepath : (str) Path where you want to save all the generated files and folders

Example

Data Load and Process Phase

import panclassif as pc 

names = pc.(featSelecthomepath,cancerpath,normalpath, k=5)
pc.dataProcess(homepath,names,cancerpath,smoothed_cancer,smoothed_normal)
pc.upsampled(names, homepath)
pc.binary_merge(names, homepath)
pc.multi_merge(names, homepath)

Classification Phase

from sklearn.ensemble import RandomForestClassifier
pc.classification(homepath, RandomForestClassifier(n_estimators=100), mode="multi", save_model="RF")

Gene enrichment check

pc.gsea(homepath)