Integrative approaches for predicting protein function and prioritizing genes for complex phenotypes using protein...

Area of Science:

• Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Biotechnologies generate vast biological data, including molecular networks.
• Integrating molecular networks with other data (sequences, gene expression) is crucial for a holistic understanding of biological systems.
• Defining protein similarity within networks is key for biological network analysis.

Purpose of the Study:

• To review applications of network similarity measures in biological studies.
• To evaluate the performance of different similarity measures for specific biological problems.
• To recommend optimal similarity metrics for network analysis.

Main Methods:

• Review of similarity measures applied to biological networks.
• Focus on four key problems: protein function prediction, gene prioritization (with and without gene expression data), and RNAi experiment validation.
• Comparative analysis of diffusion kernels against direct neighbors and shortest path distance metrics.

Main Results:

• Diffusion kernels demonstrate superior performance across all evaluated tasks.
• The effectiveness of diffusion kernels was shown for protein function prediction and gene prioritization.
• Diffusion kernels proved valuable in identifying false positives and negatives in RNAi experiments.

Conclusions:

• Diffusion kernels are a highly effective metric for biological network similarity.
• These kernels should be the preferred choice over traditional methods like direct neighbors and shortest path distance.
• The integration of diverse biological data using diffusion kernels enhances biological system understanding.