Chemical Computers: Mastering Pattern Recognition and Multitasking

Chemical computers composed of enzyme networks can carry out a range of functions, including temperature measurement and substance identification, all while avoiding the need for reconstruction after each use. This adaptability resembles biological systems more than traditional digital circuits, indicating a potential merger of computing and biological processes.

In nature, living organisms contain molecular systems that continuously integrate chemical and physical signals. For instance, cells detect nutrients, hormones, and temperature variations, adjusting to survive. Researchers have attempted to create analogs of this biological flexibility for years, including efforts to form logic gates with DNA; however, most artificial systems fall short due to their simplicity, inflexibility, or scalability challenges.

In a novel approach, researcher Wilhelm Huck from Radboud University in the Netherlands focused on allowing enzymes to interact autonomously rather than scripting every chemical step, leading to complex behaviors capable of recognizing chemical patterns.

The research team developed a system utilizing seven distinct enzymes embedded in tiny hydrogel beads found in small tubes. A liquid is introduced to these tubes, injecting short amino acid chains called peptides, which function as the “inputs” for the computer. As the peptides travel through the enzymes, each enzyme endeavours to cleave the peptide at designated sites along its length. When one cleavage occurs, it alters the peptide’s structure and the available cleavage sites, thereby affecting the actions of other enzymes.

This interdependence of reactions means that enzymes form a dynamic chemical network continually evolving, yielding unique patterns for the system to analyze. “Enzymes serve as the hardware while peptides act as the software. We address novel challenges based on the input provided,” noted Lee Dongyang from Caltech, who was not part of the study.

For instance, temperature influences the reaction rates of the enzymes. Elevated temperatures can accelerate certain enzymes faster than others, modifying the output’s mixture of peptide fragments. By employing machine learning algorithms to analyze these fragments, the researchers were able to correlate fragment patterns with specific temperatures.

Different chemical reactions can take place over various timescales, giving these systems a type of “memory” for previous inputs, enabling them to identify patterns over time. For example, they can distinguish between rapid and slow light pulses, allowing for both reactive and adaptive processing of changes in input.

The outcome is a versatile, dynamic chemical computer that interprets signals akin to a living organism rather than a static chemical circuit. “The same network undertook multiple roles seamlessly, including chemical categorization, temperature sensing with an average error margin of around 1.3°C from 25°C to 55°C, pH classification, and even responding to light pulse periodicity,” Li indicated.

The researchers were astonished by the effectiveness of the compact computer, with Huck expressing hopes for future advancements that might convert optical and electrical signals directly into chemical reactions, mimicking the behavior of living cells. “We started with just six or seven enzymes and six peptides,” he remarked. “Just imagine the possibilities with 100 enzymes.”