Batch Mode TD($\lambda$ ) for Controlling Partially Observable Gene Regulatory Networks.

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• External control of gene regulatory networks (GRNs) is crucial for preventing disease states.
• Existing methods often require complex computational models and extensive simulation time.

Purpose of the Study:

• To develop a novel, efficient method for controlling partially observable GRNs.
• To obtain control policies directly from gene expression data, bypassing model inference.

Main Methods:

• Combines batch mode reinforcement learning (Batch RL) and TD() algorithms.
• Interprets time-series gene expression data as system observations.
• Derives approximate stochastic policies directly from data without internal state estimation.

Main Results:

• Achieves control solutions for GRNs with thousands of genes in seconds.
• Significantly faster than existing methods, which struggle with dozens of genes.
• Generated approximate stochastic policies demonstrate performance comparable to existing approaches.

Conclusions:

• The proposed method offers a computationally efficient and scalable solution for GRN control.
• Direct policy generation from data eliminates time-consuming model inference and simulation phases.
• This approach advances the field of precision medicine and synthetic biology by enabling rapid development of control strategies.