Related Experiment Videos
Gene network inference from incomplete expression data: transcriptional control of hematopoietic commitment
Kristin Missal1, Michael A Cross, Dirk Drasdo
1Bioinformatics Group, Department of Computer Science, University of Leipzig Härtelstrasse 16-18, D-04107 Leipzig, Germany.
Bioinformatics (Oxford, England)
|December 8, 2005
Summary
Inferring gene regulatory networks from single-cell expression data is challenging due to incomplete information. This study presents a dynamic Bayesian network approach to predict network topology and parameters, improving inference feasibility with prior knowledge and controlled initial states.
Area of Science:
- Systems Biology
- Computational Biology
- Genomics
Background:
- Gene regulatory network inference often relies on population-level expression data, which is unreliable for asynchronous cells.
- Single-cell expression analysis offers a more accurate alternative but faces challenges with incomplete data and experimental limitations.
- Determining precursor states for network inference is difficult due to experimental constraints.
Purpose of the Study:
- To develop and evaluate a strategy for inferring gene regulatory networks from incomplete single-cell expression data.
- To predict the experimental requirements for accurate network inference based on data noise, prior knowledge, and initial state accessibility.
- To enhance the feasibility of network inference by combining partial learning with expectation maximization.
Main Methods:
- Utilized dynamic Bayesian networks for gene regulatory network inference.
- Implemented a 'Partial Learning' approach using experimental observations for network topology.
- Employed expectation maximization for refining network parameters.
- Validated the strategy through extensive computer simulations, including hematopoietic stem cell commitment and random networks.
Main Results:
- The proposed strategy effectively infers gene regulatory networks from incomplete expression data.
- Network inference feasibility significantly improves with increased ability to control initial states.
- Prior knowledge and noise reduction substantially enhance the accuracy of network inference.
- The study provides predictions for the number of experiments needed based on various parameters.
Conclusions:
- Dynamic Bayesian networks offer a robust framework for gene regulatory network inference from incomplete single-cell data.
- Controlling initial states and incorporating prior knowledge are critical for successful network inference.
- The developed strategy and its predictions can guide experimental design for more efficient and accurate network reconstruction.