Network Archaeology

Network Archaeology reconstructs the evolutionary growth history of protein-protein interaction (PPI) networks using probabilistic, likelihood-based models to infer ancestral network states.

Key Features:

• Likelihood-Based Reconstruction: Employs a likelihood-based method to infer previous states of PPI networks while preserving node identities for tracking individual proteins.
• Model Comparison and Validation: Compares growth models including duplication-mutation with complementarity, forest fire, and preferential attachment, finding duplication-based models better capture PPI evolution characteristics.
• Estimation of Protein Ages: Estimates protein ages in the yeast PPI network that align with sequence-based age assessments and structural features of protein complexes.
• Prediction of Interaction Dynamics: Models arrival times of extant and ancestral interactions and predicts extensive network rewiring with new edges forming predominantly within existing complexes.
• Duplication Rate Distribution and Network Clustering: Infers a distribution of per-protein duplication rates and tracks changes in the network clustering coefficient over evolutionary time.
• Paralogous Relationship Prediction: Predicts paralogous relationships among extant proteins to complement sequence-based inferences.
• Parameter Inference for Evolutionary Events: Infers plausible evolutionary model parameters and computes relative probabilities of different evolutionary events.

Scientific Applications:

• Reconstruction of ancestral PPI networks: Reconstructs ancestral PPI networks to infer the functional organization of ancient organisms.
• Mechanistic studies of network evolution: Assesses the roles of gene duplication and mutation in shaping network evolution.
• Evolutionary pattern identification and prediction: Identifies evolutionary patterns and predicts structural changes within protein complexes over time.
• Integration with sequence and structural analyses: Complements sequence-based analyses with structural data to improve the accuracy of evolutionary inferences.

Methodology:

Applies probabilistic models to simulate backward growth of networks; uses likelihood-based algorithms to infer historical states while maintaining node identities; compares multiple growth models (duplication-mutation with complementarity, forest fire, preferential attachment) to identify models that replicate observed network characteristics; integrates sequence and structural data to validate reconstructed histories.