Link to the code: brain-emulation GitHub repository

First Artificial Neuron to Operate at Biological Voltage Uses Bacterial Protein Nanowires

For decades, one of the core engineering problems in brain-computer interfaces has been a voltage mismatch. Biological neurons fire at roughly 70 to 130 millivolts. Most artificial neurons, built from conventional semiconductors, have required 0.5 volts or more to operate, roughly 5 to 10 times higher. That gap forces every interface between silicon and tissue to include signal amplification hardware, which adds bulk, heat, and complexity, and introduces noise that corrupts the very signals researchers are trying to read or write.

In October 2025, a team at the University of Massachusetts Amherst closed that gap. Shuai Fu, a graduate student in the Department of Electrical and Computer Engineering, and Jun Yao, associate professor, published results in Nature Communications describing an artificial neuron that operates at approximately 0.1 volts, matching the biological range while consuming 100 times less power than standard circuit designs. The paper is titled “Constructing artificial neurons with functional parameters comprehensively matching biological values.”

The Bacterial Nanowire Approach

The key material is not silicon or conventional metal oxide. It is protein nanowires synthesized by Geobacter sulfurreducens, a soil bacterium studied for its unusual ability to transfer electrons externally through conductive filaments. These pilin-based nanowires are biocompatible by construction, flexible, and capable of switching electrical resistance states at the millivolt scale.

The artificial neuron is built around a memristor, a variable-resistance component, made from these bacterial protein nanowires. When voltage across the memristor reaches the biological threshold, around 60 millivolts at 1.7 nanoamps, it switches on. A capacitor in the circuit charges and releases, producing a brief voltage spike. The accumulated charge then forces the memristor back off, creating a refractory period before the next potential spike. The result is a self-resetting cycle that closely mimics the action potential dynamics of biological neurons, without any external amplification.

Yao summarized the efficiency gain directly: “Previous versions of artificial neurons used 10 times more voltage, and 100 times more power, than the one we have created. Ours register only 0.1 volts, which is about the same as the neurons in our bodies.”

Chemical Modulation

One of the more significant aspects of the design is that it does not only produce voltage spikes. It also responds to neurochemicals through graphene mesh sensors integrated into the device. When sodium ions are present, they accelerate the memristor’s reset cycle, increasing firing frequency. Dopamine produces a dose-dependent bidirectional response: lower concentrations and higher concentrations push firing rates in opposite directions, a behavior that mirrors the complex modulatory role dopamine plays in biological neural circuits.

This capacity for neuromodulation is not a minor feature. Most artificial neurons in the literature produce spikes at fixed rates or with only voltage control. A device that integrates chemical signals into its firing dynamics is structurally closer to how neurons actually regulate activity in tissue. For any future interface that needs to both read and participate in neural computation rather than simply stimulate, the ability to respond to the local chemical environment is a prerequisite.

Communicating with Living Heart Cells

The team validated the device by interfacing it with cardiomyocytes, heart muscle cells wrapped around graphene mesh sensors. The artificial neuron processed the electrical signals from the beating cells in real time, without requiring amplification between the biological and electronic layers. The voltage levels at the interface remained within the range both systems could exchange natively.

This demonstration is deliberately conservative. Heart cells produce rhythmic electrical signals with well-characterized timing, making them a tractable first target. The broader implication is that the same architecture could interface with neurons in neural tissue, where signals are less regular but operate at the same voltage scale.

The significance is not just biocompatibility of the material, though that matters. It is that eliminating the amplification stage removes a structural bottleneck that has limited every previous attempt at tight biological-electronic coupling. Optical methods for neural stimulation avoid electrical mismatch but require genetic modification of target cells. Conventional electrode arrays interface electrically but indirectly, with impedance mismatch generating noise floors that limit single-neuron resolution. The bacterial nanowire memristor approach is, as the ScienceAlert coverage noted, a way of “whispering” to cells rather than shouting at them.

TRL and Where This Stands

This work sits at TRL 3 to 4. The fundamental principles have been demonstrated, and the device has been validated in a controlled laboratory setting with living cells. The next steps involve testing in neural tissue specifically, characterizing long-term stability of the protein nanowires in biological environments, and scaling from single-device demonstrations to addressable arrays.

The foundational material work was established in a 2020 paper from the same lab showing that Geobacter sulfurreducens protein nanowires could drive diffusive memristor switching at 40 to 100 millivolts, published in Nature Communications (DOI: 10.1038/s41467-020-15794-9). The 2025 paper extends that foundation to a complete artificial neuron with neuromodulatory response and living-cell interfacing.

For reference, current commercial brain-computer interfaces, Neuralink’s clinical trials and Paradromics’ Connexus BCI, still rely on amplification electronics between electrode arrays and signal processing hardware. The voltage mismatch problem this work solves is precisely the kind of fundamental constraint that limits how close those systems can get to individual neuron-level communication.

Implications for Neural Interface Design

The practical applications the team describes include wearable health monitoring, sweat-powered electronics that harvest ambient biological energy, and bio-inspired computing systems that use biological voltage dynamics natively. These are near-term applications. The longer-range question for brain-computer interface research and whole brain emulation is whether a device that already operates at biological voltage, responds to neurochemicals, and can interface with beating cells without amplification can eventually be embedded within neural circuits rather than just adjacent to them.

Flexible neural implants address the mechanical mismatch between rigid electronics and soft brain tissue. The UMass work addresses the electrical mismatch. Both problems need solutions before any bidirectional neural interface can operate at the scale and resolution that whole brain emulation would require. The two research directions are converging.

The research was funded by the Army Research Office, the U.S. National Science Foundation, the National Institutes of Health, and the Alfred P. Sloan Foundation.

Official Sources