Search: for  Help


Welcome to MARIMBA, the Molecular Annotation Resource for Integrating Microarrays with Bayesian Analysis.


MARIMBA is a web-based pipeline to model biological pathways using Bayesian networks. Bayesian networks (BNs) are graphical models that describe causal or apparently causal interactions between variables. The Bayesian networks are ideal for merging pathway models and high-throughput data due to the BNs' flexibility, robustness to error, and human interpretability. Compared to other modeling methods, Bayesian networks can be learned automatically by searching through large spaces of network topologies and retaining the most significant top scoring networks. Our pipeline also allows a systematic analysis and expansion of the existing pathways to identify novel hidden or unknown network regulators.

Driving Questions:

Pathway models and microarray gene expression data are often the first resources used when searching for biological mechanisms. Many questions can be addressed by these resources, such as:

  1. Can gene expression data be used to successfully reconstruct a known pathway?
  2. What other hidden molecular factors are involved in the selected pathway and dataset?
  3. Can interactions and novel regulatory mechanisms be identified among multiple pathways?
  4. How can prior knowledge from existing knowledge sources help refine pathway models and/or be used as validation?

Features and Tools in MARIMBA:

MARIMBA integrates gene expression data from publicly available databases (e.g., NCBI GEO and AfCS) or user-defined data with existing pathway knowledge (e.g., KEGG Pathways). MARIMBA provides a user-friendly graphical interface environment to simplify dataset selection, probeset/gene inclusion, observational file processing, settings selection, Bayesian network execution, consensus network comparison, iterative BN+1 expansion, and results visualization. MARIMBA also offers automated data processing tools including foldchange and clustering. MARIMBA allows users to store and update their own data and modeling results. New algorithms such as BN+1 have been developed to address fundamental pathway-related questions. MARIMBA is implemented using a PHP interface, MySQL database, and Python.

Recent MARIMBA-related publications:

Hodges AP et al. (2010) Bayesian network expansion identifies novel ROS and biofilm regulators. PLoS One 5(3):e9513.

Hodges AP, Woolf P, He Y. (2010). BN+1 Bayesian Network Expansion for Identifying Molecular Pathway Elements. Communicative and Integrative Biology. In press.


Thank you for using MARIMBA. Your suggestions and comments are welcome and appreciated!

MARIMBA is optimized in Firefox 3