In our new paper, we propose a different approach: biological computationalism. The label is meant to be provocative, but also to sharpen the conversation. Our main argument is that the standard computational framework is broken, or at least poorly suited to how brains actually work. For a long time, it has been tempting to picture the mind as software running on neural hardware, with the brain “computing” in roughly the way a conventional computer does. But real brains are not von Neumann machines, and forcing that comparison leads to shaky metaphors and fragile explanations. If we want a serious account of how brains compute, and what it would take to build minds in other substrates, we first need a broader definition of what “computation” can be.

Biological computation, as we describe it, has three core features.

Hybrid Brain Computation in Real Time

First, biological computation is hybrid. It mixes discrete events with continuous dynamics. Neurons fire spikes, synapses release neurotransmitters, and networks shift through event-like states. At the same time, these events unfold within constantly changing physical conditions such as voltage fields, chemical gradients, ionic diffusion, and time-varying conductances. The brain is not purely digital, and it is not simply an analog machine either. Instead, it works as a multi-layered system where continuous processes influence discrete events, and discrete events reshape the continuous background, over and over, in an ongoing feedback loop.

Why Brain Computation Cannot Be Separated by Scale

Second, biological computation is scale-inseparable. In conventional computing, it is often possible to cleanly separate software from hardware, or a “functional level” from an “implementation level.” In the brain, that kind of separation breaks down. There is no neat dividing line where you can point to the algorithm on one side and the physical mechanism on the other. Cause and effect run across many scales at once, from ion channels to dendrites to circuits to whole-brain dynamics, and these levels do not behave like independent modules stacked in layers. In biological systems, changing the “implementation” changes the “computation,” because the two are tightly intertwined.

Metabolism and Energy Constraints Shape Intelligence

Third, biological computation is metabolically grounded. The brain operates under strict energy limits, and those limits shape its structure and function everywhere. This is not just an engineering detail. Energy constraints influence what the brain can represent, how it learns, which patterns remain stable, and how information is coordinated and routed. From this perspective, the tight coupling across levels is not accidental complexity. It is an energy optimization strategy that supports robust, flexible intelligence under severe metabolic limits.

The Algorithm Is the Substrate

Taken together, these three features point to a conclusion that can feel strange if you are used to classical computing ideas. Computation in the brain is not abstract symbol manipulation. It is not simply about moving representations around according to formal rules while the physical medium is treated as “mere implementation.” In biological computation, the algorithm is the substrate. The physical organization does not just enable the computation, it is what the computation consists of. Brains do not merely run a program. They are a specific kind of physical process that computes by unfolding through time.

What This Means for AI and Synthetic Minds

This view also exposes a limitation in how people often describe modern AI. Even powerful systems mostly simulate functions. They learn mappings from inputs to outputs, sometimes with impressive generalization, but the computation is still a digital procedure running on hardware built for a very different style of computing. Brains, by contrast, carry out computation in physical time. Continuous fields, ion flows, dendritic integration, local oscillatory coupling, and emergent electromagnetic interactions are not just biological “details” that can be ignored while extracting an abstract algorithm. In our view, these are the computational primitives of the system. They are the mechanisms that enable real-time integration, resilience, and adaptive control.

Not Biology Only, But Biology Like Computation

This does not mean we think consciousness is somehow restricted to carbon-based life. We are not arguing “biology or nothing.” Our claim is narrower and more practical. If consciousness (or mind-like cognition) depends on this kind of computation, then it may require biological-style computational organization, even if it is built in new substrates. The key issue is not whether the substrate is literally biological, but whether the system instantiates the right kind of hybrid, scale-inseparable, metabolically (or more generally energetically) grounded computation.

A Different Target for Building Conscious Machines

That reframes the goal for anyone trying to build synthetic minds. If brain computation cannot be separated from how it is physically realized, then scaling digital AI alone may not be enough. This is not because digital systems cannot become more capable, but because capability is only part of the puzzle. The deeper risk is that we may be optimizing the wrong thing by improving algorithms while leaving the underlying computational ontology unchanged. Biological computationalism suggests that building truly mind-like systems may require new kinds of physical machines whose computation is not organized as software on hardware, but spread across levels, dynamically linked, and shaped by the constraints of real-time physics and energy.

So if we want something like synthetic consciousness, the central question may not be, “What algorithm should we run?” It may be, “What kind of physical system must exist for that algorithm to be inseparable from its own dynamics?” What features are required, including hybrid event-field interactions, multi-scale coupling without clean interfaces, and energetic constraints that shape inference and learning, so that computation is not an abstract description layered on top but an intrinsic property of the system itself?

That is the shift biological computationalism calls for. It moves the challenge from finding the right program to finding the right kind of computing matter.

Candida auris is especially dangerous for people who are already critically ill, leaving hospitals highly exposed to outbreaks. Although the fungus can exist on the skin without causing symptoms, patients who depend on ventilators face a much greater risk. Once infection occurs, about 45 percent of patients die, and the fungus is resistant to all major types of antifungal drugs. This resistance makes treatment extremely challenging and allows the pathogen to persist in hospital wards.

A global health threat with mysterious origins

The infection was first detected in 2008, and its origin is still unknown. Since then, outbreaks have been reported in more than 40 countries, including the UK. Candida auris, also called Candidozyma auris, is now recognized as a serious global health threat and appears on the World Health Organization’s critical priority fungal pathogens list. In the UK, reported cases have continued to rise steadily.

Studying infection in a living model

Researchers at the University of Exeter have now taken a major step forward by examining how genes are activated during Candida auris infection. This marks the first time such genetic activity has been studied in a living host using an approach based on fish larvae. The study was published in the Nature portfolio journal Communications Biology and was supported by Wellcome, the Medical Research Council (MRC), and the National Center for Replacement, Reduction and Refinement (NC3Rs).

The researchers say the results could help identify a biological target for new antifungal treatments or even allow existing drugs to be reused, if the same genetic behavior is confirmed during infection in humans.

The project was co-led by NIHR Clinical Lecturer Hugh Gifford of the University of Exeter’s MRC Center for Medical Mycology (CMM). He said: “Since it emerged, Candida auris has wreaked havoc where it takes hold in hospital intensive care units. It can be deadly for vulnerable patients, and health trusts have spent millions on the difficult job of eradication. We think our research may have revealed an Achilles heel in this lethal pathogen during active infection, and we urgently need more research to explore whether we can find drugs that target and exploit this weakness.”

Why traditional research models fell short

One of the biggest obstacles in studying Candida auris has been its ability to survive high temperatures. When combined with its unusually strong tolerance for salt, this has led some researchers to suggest it may have originated in tropical oceans or marine animals. These traits also made it difficult to study using conventional laboratory models.

To overcome this, the Exeter team developed a new infection model using Arabian killifish. The eggs of this species can survive at temperatures similar to the human body, making them suitable for observing infection in conditions that closely resemble real illness.

Genetic activity reveals possible vulnerabilities

During the experiments, researchers observed that Candida auris can change its shape by forming elongated fungal structures known as filaments. These structures may help the fungus search for nutrients while infecting a host.

The team also analyzed which genes were activated or switched off during infection to identify possible weak points. Several of the genes that became active are responsible for producing nutrient pumps that capture iron-scavenging molecules and transport iron into fungal cells. Because iron is essential for survival, this process may represent a critical vulnerability.

Co-senior author Dr. Rhys Farrer of the University of Exeter’s MRC Centre for Medical Mycology said: “Until now, we’ve had no idea what genes are active during infection of a living host. We now need to find out if this also occurs during human infection. The fact that we found genes are activated to scavenge iron gives clues to where Candida auris may originate, such as an iron-poor environment in the sea. It also gives us a potential target for new and already existing drugs.”

Hope for future treatments

Dr. Gifford, who also works as a resident physician in intensive care and respiratory medicine at the Royal Devon & Exeter Hospital, emphasized the clinical importance of the findings. He said: “While there are a number of research steps to go through yet, our finding could be an exciting prospect for future treatment. We have drugs that target iron scavenging activities. We now need to explore whether they could be repurposed to stop Candida auris from killing humans and closing down hospital intensive care units.”

The Arabian killifish larvae model was developed with support from an NC3Rs project grant as an alternative to using mouse and zebrafish models, which are commonly used to study interactions between pathogens and their hosts. Dr. Katie Bates, NC3Rs Head of Research Funding, said: “This new publication demonstrates the utility of the replacement model to study Candida auris infection and enable unprecedented insights into cellular and molecular events in live infected hosts. This is a brilliant example of how innovative alternative approaches can overcome key limitations of traditional animal studies.”

The paper is titled ‘Xenosiderophore transporter gene expression and clade-specific filamentation in Candida auris killifish (Aphanius dispar) infection’ and is published in the Nature portfolio journal Communications Biology.

The imaging system is designed to detect extremely weak signals from surface-enhanced Raman scattering (SERS) nanoparticles that are engineered to attach to tumor markers. Once these nanoparticles are applied to a sample or to the area being examined, the system reads their Raman signal and automatically highlights regions that are more likely to contain tumor tissue.

“Traditional methods for cancer-related diagnosis are time-consuming and labor-intensive because they require staining tissue samples and having a pathologist look for any abnormalities,” said research team leader Zhen Qiu from the Institute for Quantitative Health Science and Engineering (IQ), Michigan State University. “While our system would not immediately replace pathology, it could serve as a rapid screening tool to accelerate diagnosis.”

Published results show major gains in sensitivity

In Optica, Optica Publishing Group’s journal for high-impact research, Qiu and colleagues report that their system can distinguish cancerous cells from healthy ones while detecting Raman signals that are about four times weaker than those measured by a comparable commercial system. This improved sensitivity comes from combining a swept-source laser — which changes wavelength during analysis — with an ultra-sensitive detector called a superconducting nanowire single-photon detector (SNSPD).

“This technology could eventually enable portable or intraoperative devices that enable clinicians to detect cancers at earlier stages, improve the accuracy of biopsy sampling and monitor disease progression through less invasive testing,” said Qiu. “Ultimately, such advances could enhance patient outcomes and reduce diagnostic delays, accelerating the path from detection to treatment.”

Pushing detection limits with superconducting detectors

Qiu’s lab studies how SNSPDs can be used to enhance a range of imaging technologies. SNSPDs rely on a superconducting wire that can detect individual particles of light, allowing the system to capture extremely weak optical signals at high speed while keeping background noise very low.

For this project, the researchers aimed to build a platform that could measure Raman signals far fainter than those detected by existing Raman systems. Raman imaging works by mapping a sample’s chemical composition through the unique light-scattering fingerprints of its molecules. These signals can be strengthened by using SERS nanoparticles.

“Combining this advanced detector with a swept-source Raman architecture that replaces a bulky camera and collects light more efficiently resulted in a system with a detection limit well beyond that of comparable commercial systems,” said Qiu. “Also, the fiber coupling configuration and compact design facilitate system miniaturization and future clinical translation.”

Strong tumor contrast across multiple sample types

To test the system, the team used SERS nanoparticles coated with hyaluronan acid, which enables the particles to bind to CD44, a surface protein found on many tumor cells. Initial experiments with simple nanoparticle solutions showed that the system could reach femtomolar sensitivity. The researchers then applied the imaging platform to cultured breast cancer cells, mouse tumors, and healthy tissue samples.

“The SERS signals were strongly concentrated in tumor samples, with only minimal background detected in healthy tissue,” said Qiu. “This demonstrates both the system’s exceptional sensitivity and its ability to provide reliable tumor-versus-healthy contrast. Moreover, by adjusting or substituting the targeting molecule, this method could be adapted for other cancer types.”

Next steps toward clinical use

According to the researchers, additional work is needed before the system can be used in clinical settings. Future improvements will focus on increasing readout speed and expanding validation studies. The team is exploring faster laser sources, including VCSELs, and testing whether narrowing the sweep range can further improve performance. They also plan multiplexing experiments that use different nanoparticles to target multiple biomarkers at the same time.

The researchers acknowledge industry collaborator Quantum Opus, which provided the SNSPD devices used in this work.

Verticillin A is notoriously hard to build because of its intricate chemical architecture. Even compared with closely related compounds, it proved far more challenging to synthesize, despite differing by only a couple of atoms.

“We have a much better appreciation for how those subtle structural changes can significantly increase the synthetic challenge,” says Mohammad Movassaghi, an MIT professor of chemistry. “Now we have the technology where we can not only access them for the first time, more than 50 years after they were isolated, but also we can make many designed variants, which can enable further detailed studies.”

In lab tests using human cancer cells, one verticillin A derivative stood out against a pediatric brain cancer known as diffuse midline glioma. The researchers emphasize that additional testing is needed to assess whether it could eventually be useful in the clinic.

Movassaghi and Jun Qi, an associate professor of medicine at Dana-Farber Cancer Institute/Boston Children’s Cancer and Blood Disorders Center and Harvard Medical School, are the senior authors of the study, published in the Journal of the American Chemical Society. Walker Knauss PhD ’24 is the paper’s lead author. Xiuqi Wang, a medicinal chemist and chemical biologist at Dana-Farber, and Mariella Filbin, research director in the Pediatric Neurology-Oncology Program at Dana-Farber/Boston Children’s Cancer and Blood Disorders Center, are also authors.

Why This Fungal Molecule Was So Hard to Make

Researchers first reported isolating verticillin A from fungi in 1970. Fungi use the compound to help defend themselves from pathogens. Verticillin A and similar fungal molecules have been explored for possible anticancer and antimicrobial activity, but their complexity has made them difficult to synthesize.

In 2009, Movassaghi’s lab reported the synthesis of (+)-11,11′-dideoxyverticillin A, a compound closely related to verticillin A. That molecule contains 10 rings and eight stereogenic centers, meaning carbon atoms that each connect to four different chemical groups. Those groups must be positioned with the correct orientation, or stereochemistry, relative to the rest of the molecule.

Even after that earlier success, verticillin A itself remained out of reach. The key difference between verticillin A and (+)-11,11′-dideoxyverticillin A is two oxygen atoms, but those additions made a major difference in how the molecule behaves during synthesis.

“Those two oxygens greatly limit the window of opportunity that you have in terms of doing chemical transformations,” Movassaghi says. “It makes the compound so much more fragile, so much more sensitive, so that even though we had had years of methodological advances, the compound continued to pose a challenge for us.”

Rethinking the Chemistry Step by Step

Both versions of the verticillin molecule are built from two identical halves that must be connected into a structure called a dimer. In the earlier synthesis of (+)-11,11′-dideoxyverticillin A, the team carried out the dimerization near the end of the process and then formed four crucial carbon-sulfur bonds.

When they tried to apply that same sequencing to verticillin A, it did not work. Adding the carbon-sulfur bonds late in the process failed to deliver the correct stereochemistry, forcing the team to redesign the entire order of steps.

“What we learned was the timing of the events is absolutely critical. We had to significantly change the order of the bond-forming events,” Movassaghi says.

The new synthesis starts from an amino acid derivative called beta-hydroxytryptophan. From there, the researchers build the structure in stages, adding chemical functional groups, including alcohols, ketones, and amides, while carefully controlling stereochemistry at each step.

To guide that control, the team introduced a group containing two carbon-sulfur bonds and a disulfide bond early in the process. Because disulfides are sensitive, they had to be “masked” by converting them into a protected pair of sulfides so the structure would not break down during later reactions. After dimerization, the disulfide-containing groups were restored.

“This particular dimerization really stands out in terms of the complexity of the substrates that we’re bringing together, which have such a dense array of functional groups and stereochemistry,” Movassaghi says.

In total, the route takes 16 steps from the beta-hydroxytryptophan starting material to reach verticillin A.

Early Tests Against Diffuse Midline Glioma

With verticillin A finally accessible, the researchers could also adjust the approach to create derivates. A Dana-Farber team tested these molecules against several types of diffuse midline glioma (DMG), a rare brain tumor with limited treatment options.

The strongest effects appeared in DMG cell lines that produce high levels of a protein called EZHIP. EZHIP influences DNA methylation and has previously been flagged as a potential drug target for DMG.

“Identifying the potential targets of these compounds will play a critical role in further understanding their mechanism of action, and more importantly, will help optimize the compounds from the Movassaghi lab to be more target specific for novel therapy development,” Qi says.

The verticillin derivatives seem to affect EZHIP in a way that increases DNA methylation, which pushes the cancer cells into programmed cell death. The most effective molecules in these experiments were N-sulfonylated (+)-11,11′-dideoxyverticillin A and N-sulfonylated verticillin A. N-sulfonylation — the addition of a functional group containing sulfur and oxygen — improves molecular stability.

“The natural product itself is not the most potent, but it’s the natural product synthesis that brought us to a point where we can make these derivatives and study them,” Movassaghi says.

Next, the Dana-Farber researchers plan to further confirm how the verticillin derivatives work, and they hope to test the compounds in animal models of pediatric brain cancers.

“Natural compounds have been valuable resources for drug discovery, and we will fully evaluate the therapeutic potential of these molecules by integrating our expertise in chemistry, chemical biology, cancer biology, and patient care. We have also profiled our lead molecules in more than 800 cancer cell lines, and will be able to understand their functions more broadly in other cancers,” Qi says.

The research was funded by the National Institute of General Medical Sciences, the Ependymoma Research Foundation, and the Curing Kids Cancer Foundation.