Recurrent neural chemical reaction networks that approximate arbitrary dynamics

Area of Science:

• Biochemistry
• Synthetic Biology
• Molecular Nanotechnology

Background:

• Biological systems utilize complex dynamics like oscillations and chaos.
• Creating synthetic chemical systems with these dynamics is challenging.

Purpose of the Study:

• To introduce a molecular recurrent neural chemical reaction network (RNCRN).
• To demonstrate RNCRNs can approximate arbitrary dynamics and exhibit biologically relevant features.

Main Methods:

• Developed a modular network of chemical neurons (RNCRN).
• Proved theoretical capacity for approximating dynamics with sufficient neurons and fast reactions.
• Trained RNCRNs to display specific dynamical features.

Main Results:

• Demonstrated RNCRNs can systematically approximate any dynamics.
• Successfully trained RNCRNs with moderate complexity to exhibit biologically important dynamics.
• Showed experimental feasibility using DNA-strand-displacement technologies.

Conclusions:

• RNCRNs offer a novel approach to designing synthetic chemical systems with complex dynamics.
• This molecular computing paradigm is experimentally viable and has implications for synthetic biology.