A New ℓ₀-Regularized Log-Linear Poisson Graphical Model with Applications to RNA Sequencing Data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression data, such as RNA-seq, is crucial for understanding biological systems.
• Inferring gene networks is essential for deciphering complex biological pathways.
• Existing graphical models face challenges in accurately capturing conditional gene dependencies.

Purpose of the Study:

• To develop a new sparse Poisson graphical model for gene network inference.
• To improve the accuracy of gene network reconstruction from RNA-seq data.
• To evaluate the proposed model's performance against existing methods.

Main Methods:

• Developed a novel -based sparse Poisson graphical model.
• Applied broken adaptive ridge-regularized log-linear Poisson regression.
• Modeled conditional gene associations using a pair-wise Markov property.
• Utilized RNA-seq gene expression count data.

Main Results:

• The proposed model generates sparse gene networks with higher accuracy.
• Empirical studies demonstrate superior performance compared to -regularized Poisson graphical models.
• Successfully illustrated the model's application on kidney renal clear cell carcinoma micro-RNA-seq data.

Conclusions:

• The new sparse Poisson graphical model offers a more accurate approach to gene network inference.
• The method effectively captures conditional gene associations from RNA-seq data.
• The model shows promise for applications in cancer genomics and systems biology.