difflearn

This is a python tool packages for differential network inference (DNI).

This package mainly contains:

• Differential network inference models:

  • Pinv;
  • NetDiff;
  • BDgraph;
  • JGL;
  • JGLCV;

• Expression profiles simulation algorithms:

  • distributions:
    • Gaussian;
    • Exponential;
    • Mixed;
  • network structures:
    • random;
    • hub;
    • block;
    • scale-free;

• Visulization tools and some useful utilities.

Requirements:

Before installation, you should:

1. install pytorch yourself according to your environment;
2. install R language and R packages as follows:
   • JGL

     install.packages( "JGL" )
     

   • BDgraph:

     install.packages( "BDgraph" )
     

   • NetDiff:

     library(devtools)
     install_git("https://gitlab.com/tt104/NetDiff.git")
     

Please note: If you have several different versions of R, you should specify the version installed with above packages with:

import os
os.environ["R_HOME"] = "your path to R"

Installation

Easily run:

pip install difflearn

Quick Start

from difflearn.simulation import *
from difflearn.models import Random,Pinv,NetDiff,BDGraph,JointGraphicalLasso,JointGraphicalLassoCV
from difflearn.utils import *
from difflearn.visualization import show_matrix
import matplotlib.pyplot as plt

data_params = {
    'p': 10,
    'n': 1000,
    'sample_n': 100,
    'repeats': 1,
    'sparsity': [0.1, 0.1],
    'diff_ratio': [0.5, 0.5],
    'parallel_loops': 1,
    'net_rand_mode': 'BA',
    'diff_mode': 'hub',
    'target_type': 'float',
    'distribution': 'Gaussian',
    'usage': 'comparison',
}


data = ExpressionProfilesParallel(**data_params)

modelrandom = Random()
modelPinv = Pinv()
modelBDgraph = BDGraph()
modelNetDiff = NetDiff()
modelJGL = JointGraphicalLasso()
modelJGLCV = JointGraphicalLassoCV()
(sigma, delta, *X) = data[0]

modelrandom.fit(X)
modelPinv.fit(X)
modelBDgraph.fit(X)
modelNetDiff.fit(X)
modelJGL.fit(X)
modelJGLCV.fit(X)


fig, axs = plt.subplots(4, 2, figsize=(7,7))


show_matrix(vec2mat(delta)[0], ax=axs[0][0], title = 'Ground Truth')
axs[0][1].set_visible(False)
show_matrix(modelrandom.delta, ax=axs[1][0], title = 'Random')
show_matrix(modelPinv.delta, ax=axs[1][1], title = 'Pinv')
show_matrix(modelBDgraph.delta, ax=axs[2][0], title = 'BDgraph')
show_matrix(modelNetDiff.delta, ax=axs[2][1], title = 'NetDiff')
show_matrix(modelJGL.delta, ax=axs[3][0], title = 'JGL')
show_matrix(modelJGLCV.delta, ax=axs[3][1], title = 'JGLCV')
plt.tight_layout()
fig.set_dpi(300)
plt.show()