A scalable matrix factorization-based algorithm to jointly infer regulatory effects and transcription factor activities developed in the context of my Master's Thesis.

Consider a setting with G genes, TF proteins (transcription factors), and n samples (e.g. individual in a studies or cells). Given a gene expression matrix (matrix of dimension G x n), a prior for the regulatory network (matrix of dimension G x TF), and a protein-protein interaction network (matrix of dimension TF x TF), GIRAFFE computes:

• A matrix for the transcription factor activities TFA. Each entry describes the amount of proteins available to regulate their target genes.
• A regulatory network R of dimension G x TF. Weights can be interpreted as coefficients of a linear model that considers the transcription factor activity as covariates, and gene expression as target.

Clone the repository into your local disk:

git clone https://github.com/soelmicheletti/giraffe.git

Then install giraffe through pip:

cd giraffe
pip install -e .

Upon completion you can load netZooPy in your python code through

import giraffe

import giraffe
import numpy as np

# Generate toy data
G = 100 # Genes
TF = 20 # Transcription factors (proteins)
n = 10 # Samples (e.g. individuals)

# expression of size (G, n); prior of size (G, TF); PPI of size (TF, TF)
expression = np.random.random((G, n))
prior = np.random.randint(0, 2, size = (G, TF))
ppi = np.random.randint(0, 2, size = (TF, TF))
ppi ^= ppi.T
np.fill_diagonal(ppi, 1)

# Run GIRAFFE
giraffe_model = giraffe.Giraffe(expression, prior, ppi)

R_hat = giraffe_model.get_regulation() # Size (G, TF)
TFA_hat = giraffe_model.get_tfa() # Size (TF, n)

More details can be found in our Tutorial.

• giraffe contains the source code of our algorithm.
• notebooks/data contains the data used and generated in the experiments. Note that the version on GitHub does not contain all the data. Please download them from Zenodo.
• The jupyter notebooks in notebooks can be used to reproduce the experiments in the thesis.
• We provide an introduction to computational methods for gene regulation on Medium, with the hope to facilitate researchers without a computational biology background.

• Alexander Marx, Julia Vogt, and John Quackenbush for making this exchange possible.
• Alexander Marx, Jonas Fischer, and Panagiotis Mandros for their supervision and invaluable guidance throughout this project.
• Marouen Ben Guebila, Rebekka Burkholz, Chen Chen, Derrick DeConti, Dawn DeMeo, Viola Fanfani, Intekhab Hossain, Camila Lopes-Ramos, John Quackenbush, Enakshi Saha, Katherine Shutta, and Julia Vogt for thoughtful critiques and discussions.