Markov chain Monte Carlo for active module identification problem.

Area of Science:

• Systems biology
• Bioinformatics
• Computational biology

Background:

• Integrative network methods are crucial for interpreting high-throughput biological data.
• Current methods for identifying active modules often yield single subnetworks with uncertain vertex inclusion.
• This limitation necessitates a more nuanced approach to active module identification.

Purpose of the Study:

• To develop a method for soft classification of vertices within active modules.
• To estimate the probability of each vertex belonging to an active module.
• To address the limitations of hard classification in current active module identification methods.

Main Methods:

• Utilizing Markov chain Monte Carlo (MCMC) subnetwork sampling.
• Applying the method to estimate vertex probabilities for active module membership.
• Evaluating performance on simulated and real biological datasets.

Main Results:

• The proposed MCMC method provides reliable probability estimates for vertex inclusion in active modules.
• The method demonstrates consistency with existing resampling techniques on large datasets.
• It achieves high classification performance on datasets where other methods are inapplicable, such as those with few replicates.

Conclusions:

• The developed method enables estimation of individual vertex probabilities and false discovery rates (FDR) for active modules.
• It allows for the consistent identification of connected subgraphs at specified FDR levels.
• The approach shows strong computational performance and high classification accuracy on diverse datasets.