A tensor-based bi-random walks model for protein function prediction.

Area of Science:

• Computational biology
• Bioinformatics
• Systems biology

Background:

• Accurate protein function characterization is vital for molecular understanding, impacting biomedicine and pharmaceuticals.
• Computational protein function prediction has been extensively studied.
• Integrating multi-omics data with network topologies presents a significant challenge due to noise and errors in protein-protein interaction (PPI) networks.

Purpose of the Study:

• To develop a novel computational method for accurate protein function prediction.
• To address the limitations of existing methods in handling noisy PPI networks and integrating diverse biological data.
• To establish a robust data model for preserving intrinsic characteristics of network topologies and biological data.

Main Methods:

• Proposed the Random Walks with Restart on Tensor (RWRT) method.
• Constructed a functional similarity tensor by integrating protein interaction networks with multi-omics data (domain annotation, protein complex information).
• Extended the bi-random walks algorithm to a tensor framework for scoring protein functional similarity and applied a cohesiveness coefficient for filtering.

Main Results:

• RWRT significantly outperforms state-of-the-art methods in protein function prediction.
• Achieved at least an 18% improvement in the area under the receiver-operating curve (AUROC).
• Demonstrated effective filtering of potential false positives and accurate annotation of target proteins.

Conclusions:

• A functional similarity tensor provides an effective alternative for integrating diverse network data.
• The tensor-based random walk model successfully discovers functionally similar proteins and mitigates errors in PPI networks.
• Effective protein function prediction relies on extracting and leveraging comprehensive functional similarity information from protein correlations.