Learning in Transcriptional Network Models: Computational Discovery of Pathway-Level Memory and Effective...

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Trainability, the capacity to modify future behavior based on past experiences, is fundamental to control and prediction.
• While extensively used in computer and behavioral sciences, biological trainability is typically limited to animal behavior.
• Emerging research in basal cognition suggests memory and learning capabilities exist in non-neural biological systems.

Purpose of the Study:

• To characterize memory capabilities in biological gene regulatory circuit and protein pathway models.
• To identify stimuli-based interventions for controlling undesirable network behaviors like drug resistance.
• To explore the stability, robustness, and underlying network properties associated with biological memory.

Main Methods:

• Modeling of continuous gene regulatory networks and protein pathways.
• Analysis of network dynamics under various stimulation regimes.
• Assessment of memory stability, noise robustness, and correlation with network properties.

Main Results:

• Biological models demonstrate diverse memory capabilities, extending prior findings in binary networks.
• Specific stimulation patterns can abolish drug resistance and sensitization, offering targeted control.
• Memories exhibit long-term stability, robustness to noise, and in some cases, noise enhances memory potential.
• No single network property universally indicates memory; biological network structure is crucial, unlike random networks.

Conclusions:

• Gene regulatory circuits and protein pathways possess inherent memory, enabling 'trainability' in non-neural systems.
• Stimuli-derived control of dynamic pathways offers a novel approach to influencing cellular behavior.
• This work opens avenues for studying proto-cognitive capacities and presents potential biomedical applications as an alternative to gene therapy.