Improving GRN re-construction by mining hidden regulatory signals.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Inferring gene regulatory networks (GRNs) from gene expression data is crucial for understanding cellular mechanisms.
• Existing methods face challenges in identifying all regulatory factors, especially latent ones.

Purpose of the Study:

• To develop a dictionary learning-based approach for inferring GRNs by mining both known and latent regulatory signals.
• To improve the accuracy and completeness of GRN inference.

Main Methods:

• Modified the k-SVD dictionary learning algorithm to incorporate the sparse properties of GRNs.
• Developed a method to mine regulatory signals, including hidden ones, from gene expression data.
• Calculated confidence scores for transcription factor-target gene interactions.

Main Results:

• Successfully recovered hidden regulatory signals in simulated data.
• Demonstrated superior performance of the proposed algorithm (OURM) over state-of-the-art methods (GENIE3, ARACNE) on real-world datasets.
• Achieved higher area under the receiver operating characteristic curves (AUROC) and area under the precision-recall curves (AUPRC).

Conclusions:

• The proposed dictionary learning approach effectively infers gene regulatory networks by uncovering latent regulatory factors.
• OURM offers a significant advancement in GRN inference accuracy and completeness.