Are scale-free networks robust to measurement errors?

Area of Science:

• Network science
• Systems biology
• Bioinformatics

Background:

• Scale-free networks are common in complex systems, including biological networks.
• Existing studies often overlook false positive and false negative links in network data.
• Measurement errors significantly impact the topological inference of biological networks.

Purpose of the Study:

• To investigate the impact of erroneous links on network topological inference.
• To explore error mechanisms in scale-free networks, focusing on yeast protein interaction networks.
• To identify robust methods for estimating network parameters in the presence of errors.

Main Methods:

• Theoretical derivations and simulations were employed to analyze network topology.
• Comparison of real yeast protein interaction data with simulated data to understand error mechanisms.
• Evaluation of different robust estimators for scale parameter estimation.

Main Results:

• Ignoring erroneous links can bias scale parameter estimates; robust estimators are recommended.
• Scale-free property holds for mid-range connectivities but is distorted at low/high connectivities with errors.
• Different error mechanisms (false positives/negatives) affect yeast protein interaction networks from various data sources.

Conclusions:

• Robust estimators, like least trimmed mean squares, are crucial for accurate scale parameter estimation.
• The scale-free property's robustness varies depending on the error mechanism.
• Yeast protein interaction networks exhibit distinct error profiles (e.g., high false negatives in MIPS, variable rates in yeast two-hybrid data).