Related Experiment Videos
Sensitivity and specificity of inferring genetic regulatory interactions from microarray experiments with dynamic
1Biomathematics and Statistics Scotland, JCMB, The King's Buildings, Edinburgh, EH9 3JZ, UK. dirk@bioss.ac.uk
Bioinformatics (Oxford, England)
|November 25, 2003
Summary
This study validates Bayesian networks for inferring gene regulatory networks from gene expression data. Simulation results show performance depends on data size and prior knowledge, offering a reliable method for systems biology.
Area of Science:
- Computational Biology
- Systems Biology
- Bioinformatics
Background:
- Inferring genetic regulatory interactions from gene expression data is challenging due to small datasets.
- Previous validation methods using literature are unreliable due to lack of gold standards.
Purpose of the Study:
- To test the viability of Bayesian networks for inferring gene regulatory networks using a realistic simulation study.
- To evaluate the performance of Bayesian networks in reverse engineering biological networks.
Main Methods:
- Simulated gene expression data from a realistic biological network (DNA, mRNA, proteins).
- Inferred interaction networks using Bayesian networks and Bayesian learning with Markov chain Monte Carlo.
- Assessed performance using receiver operator characteristics (ROC) curves.
Main Results:
- Simulation results presented as ROC curves, enabling estimation of true and false interaction detection rates.
- Network inference performance is sensitive to training set size and prior assumptions.
- Performance is influenced by experimental sampling strategy and inclusion of sequence-based information.
Conclusions:
- Bayesian networks offer a viable approach for inferring gene regulatory networks.
- Simulation studies provide a reliable framework for evaluating network inference methods.
- Understanding factors influencing performance is crucial for accurate biological network reconstruction.