DNA computing uses DNA molecules as information-bearing components in physical computations. Researchers program sequences and molecular interactions so that binding, displacement, concentration changes or self-assembly implement logic and transform inputs into outputs.
This page covers the molecular branch of biocomputing. Most of Biocomputers focuses on living neural systems. Both branches use biological molecules or tissues directly in information processing.
What changed recently
In 2025, Kevin Cherry and Lulu Qian demonstrated supervised learning in a DNA neural network. The molecular system learned to classify 100-bit patterns from training examples, with learned information stored in molecular concentrations.
A second 2025 Nature paper showed heat-rechargeable DNA logic circuits and neural networks. Heating and cooling reset the molecular system for repeated rounds of computation.
On 16 September 2026, Tristan Stérin and colleagues reported a Scaffolded DNA Computer that computes while relaxing towards a thermodynamically favoured equilibrium. The paper demonstrated ten programs, including multiplication, parity detection and addition of 25-bit numbers, a 100-bit computation. Small instances completed in under a minute, and some programs were reused dozens of times.
Computing towards equilibrium
Many molecular circuits rely on carefully controlled kinetics and spend chemical free energy keeping unwanted states suppressed. The Scaffolded DNA Computer uses an energy landscape in which correct outputs are strongly favoured at equilibrium. Errors can be displaced as the system relaxes towards the target configuration.
DNA computers currently address specialised problems with laboratory handling, slow physical steps and very different I/O constraints from silicon systems. The 2026 work demonstrates a programmable route to molecular-scale computation in which reuse and error tolerance emerge from the physical design.
How DNA computing differs from neural biocomputing
DNA computers usually encode information in molecular species, concentrations and binding configurations. Neural biocomputers use the electrical dynamics and plasticity of living neuronal networks. DNA systems excel at enormous molecular parallelism and direct interaction with chemical environments; neuronal systems are being explored for adaptation, temporal dynamics and closed-loop control.
The two branches may eventually overlap in hybrid biological systems. Their present hardware, timescales and benchmark problems are quite different.