MRFSeq

MRFSeq integrates RNA-Seq read counts with gene coexpression data using a Markov Random Field to improve differential gene expression inference, particularly for low read count genes.

Key Features:

• Markov Random Field Model: Employs an MRF framework to combine RNA-Seq read counts with gene coexpression information for differential expression inference.
• Bias Reduction: Integrates coexpression data to reduce bias against low read count values in differential expression estimation.
• Clique Potential Functions: Uses a strategic selection of clique potential functions that transforms maximum a posteriori estimation into a maximum flow problem solvable in polynomial time.
• Improved Sensitivity: Demonstrated higher sensitivity on simulated and real RNA-Seq datasets, increasing detection for low read count genes from 11.6% to 38.8% in the MAQC dataset.
• Implementation: Implemented in C for efficient computation.

Scientific Applications:

• Systems biology: Provides more reliable differential expression results to support systems-level analyses.
• Differential gene expression analysis: Improves accuracy of DE inference across a range of gene expression levels, especially for low-count genes.
• Gene regulatory network analysis: Enhances detection of expression changes relevant to reconstructing and interpreting gene regulatory networks and biological processes.

Methodology:

The algorithm integrates RNA-Seq read counts with gene coexpression data via a Markov Random Field model and is implemented in C.