XMRF

XMRF fits Markov networks to high-throughput genomics data to model interactions among genes, mutations, copy number variation, and methylation for discovery of disrupted disease networks.

Key Features:

• Modeling Versatility: Implements exponential family Markov Random Fields (ERF-MRFs) to accommodate RNA-sequencing, mutation, copy number variation, and methylation data, using Poisson graphical models for RNA-sequencing count data, Ising models for mutation and copy number categorical data, and Gaussian graphical models for methylation continuous data.
• Native Distribution Modeling: Leverages the native distribution of each dataset type rather than assuming Gaussian distributions to improve the accuracy of network structure learning.
• Efficient Computation: Incorporates parallelization techniques within its algorithms to enable efficient computation of large-scale biological networks.

Scientific Applications:

• Genetic Network Learning: Fits Markov networks to elucidate genetic interactions and pathways from RNA-sequencing, mutation, copy number variation, and methylation datasets.
• Disease Mechanism Exploration: Models disrupted networks to provide insights into molecular underpinnings of diseases and generate hypotheses for therapeutic investigation.

Methodology:

Uses the exponential family Markov Random Fields (ERF-MRFs) framework (Yang et al., 2012), including Poisson graphical models, Ising models, and Gaussian graphical models.