IEEE Biomedical News

• How a Google DeepMind Spin-off Hunts Hidden Drug Targets

  

  For more than a decade, artificial intelligence has been touted as a way to dramatically accelerate drug discovery. Yet despite billions of dollars in investment, relatively few AI-designed medicines have made it to patients. That’s partially because the timelines for careful drug testing can’t be easily compressed—and partially because drug development is just really hard.

  Isomorphic Labs, the Google DeepMind spin-off that’s building on DeepMind’s Nobel Prize-winning work on protein structure prediction, may be making the most progress. The company has signed major drug-discovery partnerships with Novartis and Eli Lilly and recently raised US $2.1 billion in funding. In February, it published a technical report describing its new Isomorphic Drug Design Engine, a system created to discover the “pockets” on proteins where drugs can bind and in general to predict how proteins and drug molecules interact.

  IEEE Spectrum spoke with Adrian Stecuła, a group leader in the machine learning organization at Isomorphic Labs, about how close AI may be to becoming a practical tool for designing new medicines.

  Going Beyond AlphaFold

  AlphaFold2 and AlphaFold3 were massive leaps forward for computational biology. Why weren’t those models sufficient for actually designing drugs?

  Adrian Stecuła: AlphaFold2 was eventually recognized with the Nobel Prize, because it arguably solved the problem of protein folding. But proteins don’t exist in a vacuum, right? They interact with a wide variety of other types of biomolecules, which involves nucleic acids, small molecule ligands, ions, and other proteins. AlphaFold3 introduced a way to model the rest of these cellular biomolecules as part of a single framework. So all of a sudden, we have a single model that can model all of these interactions all at the same time.

  That said, in the years since the AF3 release, multiple groups have evaluated it along the axis of pocket novelty. And you could see that as the pocket distance grows away from the training set, the model performance decreases. So if you define the success as “how well did the model actually fold this particular ligand with this particular protein,” as those systems become more novel, you can see a decline in performance.

  But for drug discovery, ideally we do want to pursue novel mechanisms of action, which might involve targeting a never-before-observed pocket. And so it is absolutely important for us to have our models generalize to these regions that are distant from training.

  How does theIsomorphic Drug Design Engine (IsoDDE) address these limitations, and what exactly is it predicting?

  Stecuła: It takes a lot more than just structure prediction to create a molecule that will ultimately become a drug. You don’t just need to predict where the ligand binds with the protein, but also potentially how it binds, how tightly it binds, and a plethora of other properties about the ligand and how the ligand interacts with the rest of the proteins in the body.

  IsoDDE is a unified computational system that lends itself to a number of different endpoints. As part of the technical report about IsoDDE, we have described three of those endpoints, which are structure prediction, pocket identification, and binding affinity prediction. [Editor’s note: Binding affinity measures how strongly a molecule binds to a target protein.]

  Finding Hidden Protein Pockets

  In your technical report about IsoDDE, you highlighted a key example involving a protein called cereblon and its “cryptic pocket.” First, can you tell readers about this protein and what a cryptic pocket is?

  Stecuła: Cereblon is one of the most important and well-studied proteins in the targeted protein degradation pathway. It’s part of the mechanism that’s responsible for degrading proteins in the cell. [Editor’s note: Some drugs use cereblon to mark disease-causing proteins for destruction by the cell.]

  And cryptic pockets are pockets on protein surfaces that are non-obvious, in that in the [unbound] state of the protein, meaning that if you were to look at the protein by itself, it would not have a cavity there. This pocket only opens upon the binding of just the right ligand. So you can think of it as: You need the perfect key to unlock this lock.

  How did you use recently published findings about cereblon to validate IsoDDE?

  Stecuła: In January of this year a Nature paper published a completely novel, never-before-observed cryptic pocket on the surface of this protein. First we asked the question: Can IsoDDE find this pocket just using the protein sequence as input? And we were able to perfectly predict the location of this cryptic pocket. Again, note that this pocket had never been disclosed before.

  The second question was: Can IsoDDE accurately predict how the ligands bind to the protein? Are we able to recapitulate the crystal structure shown in theNature paper? And the model was able to place both the orthosteric ligand [at the known binding site] as well as the allosteric ligand [at the new, cryptic pocket] in exactly the perfect locations.

  Most drugs today are small molecules—relatively simple compounds that bind to proteins. Does IsoDDE expand the toolkit for tackling diseases?

  Stecuła: I think many of the hopes for machine learning for drug design revolve around making more protein targets tractable. There are already many diseases that have known associated proteins. So we know that if only we could target a particular protein, we would have a chance at helping the patient population suffering from a particular disease.

  But in many of those instances, the protein that is associated with that disease doesn’t have a pocket or mechanism that can be easily drugged. IsoDDE is enabling us to find those mechanisms. Further, these methods generalize not just to small molecules but also to antibodies, molecular glues, and peptides. It’s not just a breakthrough that will impact small molecule design, but these other therapeutic modalities will also benefit from this.

  There is a lot of hype around AI in drug discovery. What do people commonly misunderstand about where the field is right now?

  Stecuła: Perhaps the misunderstanding is that just because we are able to accurately model structure, that drug discovery is a solved problem. It does take, we believe, a unified system such as IsoDDE with a plethora of other endpoints to really model these systems. We will continue to improve our performance on the endpoints that we have disclosed as part of the IsoDDE report, and will also continue to push on the endpoints that we have not yet disclosed.

  Do you imagine more of the drug discovery process becoming automated, with AI systems generating hypotheses, testing hypotheses, and analyzing results?

  Stecuła: Absolutely. This was quite nicely framed by our president Max Jaderberg as part of his TED AI talk, where he was discussing the future of agentic workflows in drug discovery. I absolutely think that is part of our collective future.

• Ultrasound Patch Could Form Future Pacemaker

  

  With the reliability of a quality wristwatch,pacemakers send out electric pulses to keep your heart beating at a steady rate. But unlike a watch, when the batteries need replacement, it’s a surgical affair—one that can be requiredas often as every five years. While the risks of having a pacemaker implanted are low, going under the knife always creates the potential for complications.

  A group of California and Massachusetts scientists have developed a pacemaker that works without requiring surgery, publishing their work last week in Nature Biomedical Engineering. Researchers developed a wearable alternative to the traditional pacemaker measuring in at about the size of an iPod Shuffle. The device sticks to the patient’s chest, putting out ultrasound waves that tell the heart to beat.

  But the ultrasound technology is only one component of what makes the noninvasive pacemaker work. Patients would undergo a gene therapy procedure to help heart cells react to the high frequency waves. Delivered by a simple injection, the treatment would work like a signal booster for the waves of the ultrasound device.

  The approach has proved effective in rats, pig hearts—used for their similarity to human hearts—and samples of human heart cells.

  “This is a very innovative and exciting study,” says Dr. Roger J. Hajjar, who heads the Gene and Cell Therapy Institute at Mass General Brigham and was not involved in the work. In terms of demonstrating that the technology could be safe and effective, he says, “the work is impressive.”

  How an ultrasound pacemaker would work for patients

  In a clinical setting, treatment would begin with a gene therapy injection that helps heart cells “hear” the ultrasound signals. It works by prompting the cells to produce a sound-sensitive protein in the ion channels that dot their membranes. The broad term for this type of therapy is “sonogenetics”—priming cells to respond to sound. It’s the same idea as a type of gene therapy that makes cells react to light, called optogenetics, which has been studied for treating hearing impairment and pain.

  Importantly, this gene therapy doesn’t alter DNA, says Gengxi Lu, who was a mechanical engineering postdoctoral researcher at MIT while working on the paper and is now a senior ultrasound engineer at Meta. Rather, he says, the treatment introduces RNA into cells that directs them to create the ultrasound-sensitive protein without altering their genetic code.

  A wearable device would then be stuck to the patient’s chest like a bandage. The device connects by wire to a data and power module that can be placed in a pants pocket. It would look similar to an insulin pump.

  The ultrasound patch is programmed to emit high frequency waves that the cell proteins receive. These signals stimulate the cells’ ion channels to let in calcium. This influx of calcium ions cues the heart to beat.

  Is sonogenetics a viable solution?

  In terms of future research, sonogenetics is ripe for solving health issues, says Chen Gong, who was a Ph.D student at the University of Southern California while conducting the research and is now a postdoctoral researcher at MIT. Researchers from Harvard; the University of California, Los Angeles; and Caltech also contributed to the research.

  Source images: Chen Gong, Qifa Zhou, et al.

  Using the sonogenetics technology, “we can modulate almost anywhere we want to stimulate, like in the inside of brains, eyes, other organs,” he says.

  But the ultrasound pacemaker still needs to prove itself. For the device to be clinically practical, Dr. Hajjar says, it would have to be comparably reliable to traditional pacemakers under limiting factors like exercise, long-term use, and anatomical differences. Additionally, gene therapy, though FDA approved in some contexts, is a hurdle in terms of costs, safety, and regulation.

  “The biggest question is not whether the ultrasound device works—it appears quite promising—but whether the benefits of a noninvasive pacing system are large enough to justify exposing patients to cardiac gene therapy when current pacemakers already have excellent safety and performance,” he says. “That will ultimately determine the size of the clinical opportunity.”

• Could This Blood-Filtering Device Help Treat Ebola?

  

  As the deadly Bundibugyo strain of Ebola continues to ravage parts of Central Africa, physicians once again find themselves scrambling for ways to keep the sickest patients alive.

  Existing antibody treatments are strain-specific and don’t target the virus responsible for the current outbreak, leaving few therapies capable of clearing virus from the bloodstream. This forces doctors to rely largely on supportive care for people in advanced stages of disease.

  That treatment gap is reviving interest in experimental blood-filtering devices that can physically remove viral particles from the bloodstream.

  These systems have been studied primarily as treatments for cancer, where they help remove tumor-derived particles, and in more common infectious diseases such as COVID-19 and hepatitis C. However, one such device was successfully deployed during the last major outbreak of Ebola, helping to drive down exceedingly high viral levels in one critically ill patient.

  If the current outbreak expands further, as some infectious-disease experts warn it could, the same technology may once again be called into action—not just as a desperate last-resort intervention for a single patient, but as a potential tool for keeping more Ebola patients alive.

  “It could really help,” says Stefan Büttner, a nephrologist and intensive-care specialist at the Klinikum Aschaffenburg-Alzenau in Germany.

  Filter Removes Millions of Viral Particles

  This year’s Ebola outbreak, though serious, is nowhere near the scale of the catastrophic epidemic that started in late 2013 and persisted for nearly 2.5 years.

  Back then, there were more than 28,000 confirmed cases and 11,000 deaths—mostly in the West African nations of Sierra Leone, Liberia, and Guinea, though the virus spread in nearby countries as well. Today, the toll is far smaller: roughly 1,000 suspected cases and fewer than 300 related deaths, all concentrated in the eastern Democratic Republic of Congo, with limited spillover into Uganda.

  Still, the emergence of the Bundibugyo strain, coupled with the lack of approved therapies designed to target it, has raised fears that doctors could once again find themselves without effective tools if the virus spreads further.The situation was similar in 2014, before the development of monoclonal antibody therapies that dramatically improved survival against the more common Zaire strain of Ebola. So, when a Ugandan doctor—infected with Ebola while treating patients in Sierra Leone—was medevacked to Germany in critical condition, the ICU team at Frankfurt University Hospital, which included Büttner at the time, tried nearly everything they could.

  Nothing seemed to halt the disease’s progression. The man’s condition only got worse. His organs began to shut down.

  Then, with emergency approval from German regulators, Büttner and his colleagues connected the patient to the Hemopurifier, a baton-size cartridge filled with sticky proteins from the common snowdrop plant. These proteins, like a kind of molecular Velcro, latch onto sugar molecules that coat viruses like Ebola and trap them as blood passes through the system.

  In this illustration, a zoomed-in view of the Hemopurifier shows how it traps the Ebola virus by passing blood through tiny fibers coated with sticky proteins.Aethlon Medical

  The Hemopurifier itself is not electrical. Instead, it connects inline to an intensive-care-grade dialysis machine, the artificial-kidney-like device that pumps the patient’s blood through its own filter to strip out toxins and surplus fluid before returning it. The Hemopurifier rides on that same circuit, and on that same machine’s electronics. The dialysis unit’s pumps push the blood through the cartridge, while its sensors balance fluid, watch circuit pressures for safety, and automatically meter the anticoagulant that keeps the blood from clotting along the way.

  Though he had been on emergency dialysis for days, the Ugandan doctor had the Hemopurifier added into the circuitry for just 6.5 hours. His blood sloshed through the device’s tiny channels and pressed against its protein snares. By the end of the brief treatment, the device had captured a whopping 253 million copies of the Ebola virus, and the man’s situation quickly turned around.

  His viral load dropped from around 380,000 particles per milliliter of blood before the procedure to roughly 6,000 the next day. His immune system, no longer overwhelmed by runaway viral replication, then regained the upper hand and finished the virus off on its own.

  Less than a week after the treatment, as Büttner’s team reported in 2015 in the journal Blood Purification, the patient was Ebola-free.

  The Technology Is Ready to Deploy

  Though it is impossible to know how much of the Ugandan doctor’s recovery could be attributed to blood filtration, Büttner believes the treatment played an important role. And he is confident the approach could prove even more beneficial for patients with much higher viral loads, who might not otherwise survive, while also helping to limit the organ damage and other complications that often arise during prolonged stays in intensive care.

  “Earlier is better,” Büttner says.

  Should the need to test that idea emerge during the current outbreak, Aethlon Medical, the company behind the Hemopurifier system, says it is prepared to move quickly.

  Back in 2014, the company secured FDA authorization for a compassionate-use protocol allowing the Hemopurifier to be used in up to 20 patients with Ebola across 10 clinical sites in the United States. More than a decade later, authorization remains active and available for use, according to the company. “That avenue is still open,” says chief medical officer Steven LaRosa.

  And although the device has never been evaluated against the Bundibugyo strain, LaRosa says its mode of action suggests it should work regardless of Ebola subtype. Given Büttner’s experience treating the man infected with the Zaire strain, together with laboratory studies demonstrating capture of the related Marburg virus, he expects the Hemopurifier would be able to filter Bundibugyo virus as well.

  “I have confidence that it would likely be removed,” LaRosa says.

  For proponents of blood filtration, the major obstacle is therefore not technological. The devices already exist, can be integrated into standard dialysis and critical-care equipment, and appear capable of capturing a broad range of pathogens, Ebola included.

  The harder challenge, they say, is convincing physicians, regulators, and health systems to embrace a treatment paradigm built around physically extracting disease-causing agents rather than targeting them with pharmaceuticals. And even if that skepticism were to fade, major logistical challenges remain.

  The Hemopurifier and other systems like it are designed to operate with dialysis-style blood-circulation systems that require specialized equipment, reliable power, trained personnel, and large-bore vascular catheters. Such resources are readily available for patients who can be evacuated to major European medical centers in places like Frankfurt. They are typically nonexistent in the austere settings where Ebola outbreaks most often occur.

  What the field still needs is therefore a “ruggedized” version of the technology that can hold up outside the controlled environment of a hospital ICU, says Michael Super, an infectious-disease researcher at the Wyss Institute for Biologically Inspired Engineering at Harvard who has spent years developing his own blood-cleansing devices.

  “That, from a practical point of view, could be something that’s very useful,” he says.

  Designing for the Outbreak Zone

  Lower-tech versions of blood-filtration systems are in development, and some medical-device makers have begun sketching out designs that could, in principle, operate without any hospital infrastructure—some even without electricity.

  For example, patent filings from Stavro Medical, a company recently acquired by ExThera, describe a manual system in which a health care worker uses syringes to push blood through a filter cartridge in batches—or, alternatively, simply raises one reservoir above another so that blood flows downhill through the filter on its own.

  Aethlon, for its part, is pursuing a more modest goal. According to LaRosa, the company is developing a stripped-down version of its Hemopurifier system that could run through a standard IV line rather than the thick catheter that dialysis often requires. “That’s not ready for prime time yet,” he says. “But we’re working on it.”

  In the end, however, what may push blood filtration into the Ebola treatment tool kit is not an engineering advance but a body count. A spreading outbreak could hasten the climb from experimental footnote to front-line tool.

• Poetry for Engineers: Cyborg Laboratory

  

  This is the place where you face yourself,the you that could be you with a fewdifferent parts, a pump for your heart,eyes off color, and fresh off the shelffake hair (a bit obvious), skin smoothed.You’re not perfect, but it’s a good start.

  Down to small digits, you’ll be improved.Memory maintained by small motors,as long as these gizmos don’t glitch.What’s before you? Full replacement ora constant game of test and switch,pieces peeled off, disconnected, removed,until you are not yourself, at least,not the self you knew. That self has ceased,bit by bit less you at each release.

• Leap in DNA Synthesis Slashes Time to Build New Genetic Sequences

  

  A new method for writing DNA promises to unlock the potential of generative AI in biology, giving scientists a fast, affordable, and accurate way to physically build the novel genetic sequences that predictive models are now producing faster than anyone can construct them.

  The technique, called Sidewinder, can assemble dozens of genetic sequences simultaneously in a single test tube, producing just one incorrect junction for every 10 million assembly events—a level of precision that far surpasses conventional methods, which misfire roughly once every 10 to 30 joins. Sidewinder also draws on cheap raw materials that have until now been too difficult to use reliably.

  “It’s a step change,” says Thomas Gorochowski, a bioengineer at the University of Bristol, in England, who was not involved in the research. “It really opens up the feasibility of synthesizing large genetic systems, maybe even small genomes.” And that, he adds, “is uber-important for all of the AI stuff that’s coming out at the moment around generative genome sequences.”

  The advance, presented earlier this month at SynBioBeta 2026 in San Jose, Calif., and detailed in a preprint posted to bioRxiv, addresses one of the more vexing mismatches in modern genomics research. Generative AI tools like Evo 2, trained on the genetic code of millions of organisms, can design new DNA sequences on demand at extraordinary speed. But physically constructing long DNA sequences in a laboratory has remained slow and expensive, especially when building not just one sequence at a time but dozens of different designs simultaneously, as testing AI predictions at scale demands.

  RELATED: Can Biologists Rewrite the Genome’s Spaghetti Code?

  In a demonstration of how squarely Sidewinder targets this bottleneck, the team behind the technique, led by Caltech synthetic biologist Kaihang Wang, harnessed the power of Evo 2 to redesign a 12,500-letter DNA sequence of the E. coli genome in silico and then used Sidewinder to build it from scratch—with no errors. Sequences of that length can encode entire biochemical pathways, laying the groundwork for engineered microbes that manufacture drugs, biofuels, or specialty chemicals, and eventually to the assembly of vast DNA constructs approaching complete artificial genomes.

  In the past, says Brian Hie, the Stanford computational biologist whose lab developed Evo 2, a project like this would likely take more than a month, based on his team’s experience with conventional commercial methods. “With a technology like this,” he says, “you could probably achieve the same thing in a few days.”

  To commercialize Sidewinder, [from left] Noah Robinson, Kaihang Wang, Adrian Woolfson, and Brian Hie cofounded a company called Genyro.Marcus Ubungen

  A New Assembly Logic

  The new method builds on a DNA synthesis strategy that Wang and his colleagues first outlined at the beginning of the year in Nature, but with substantially greater capacity.

  Thanks to a new algorithm that automates the most computationally demanding part of the process and laboratory innovations in how raw ingredients are managed, it is now feasible to synthesize ever larger and more numerous DNA constructs simultaneously. This opens up applications including drug discovery, data storage, and the design of synthetic organisms.

  “The pace at which you can start to explore these things just opened up massively,” Gorochowski says.

  To understand how Sidewinder works, it helps to understand how DNA is typically made in a laboratory. The process begins with short, chemically manufactured strands called oligonucleotides, or oligos, the molecular alphabet blocks from which longer sequences are assembled.

  Ordering oligos individually is reliable but expensive. Scientists discovered years ago that they could slash costs by synthesizing thousands of different oligos together in a single pool. But doing so creates a chaotic soup in which fragments tangle with unintended partners, leading to errors.

  Sorting out specific sequences from such a pool has traditionally required elaborate separation steps: physically dividing up the fragments, isolating them in tiny droplets, or fishing them out one by one with laser light. Each approach added cost, time, and specialized equipment.

  The Caltech team sidestepped the problem entirely.

  Page Numbers for DNA

  Sidewinder also starts with oligos, the kind anyone can buy from DNA synthesis vendors such as GenScript or Twist Bioscience, but tags each fragment with a unique molecular barcode. This short identifying sequence ensures that each piece links up only with its intended neighbor in the order that will yield the desired genetic sequence. When two bar-coded fragments meet, they form what chemists call a three-way junction: a fleeting molecular knot that locks the pieces in alignment before being cleanly removed, leaving a seamless strand.

  Wang likens these barcodes to page numbers. Whereas conventional assembly is like collating an unnumbered manuscript by matching the last line of one page to the first line of the next—workable for a short document, a recipe for chaos when sequences repeat—Sidewinder’s barcodes guide each fragment to its correct partner regardless of what sequence it carries.

  The original Sidewinder protocol required a computationally intensive calculation to design those barcodes, however, and this became impractically slow as the number of fragments grew.

  A former Caltech undergraduate student named Jean-Sebastien Paul developed a workaround. While working in Wang’s lab one summer, Paul, who is now pursuing a Ph.D. at Stanford, built a software tool called PyWinder that churns out the barcodes in minutes on a standard laptop, replacing a calculation that had previously been too slow to scale.

  Bioengineer Noah Robinson, a postdoc in Wang’s lab who codeveloped the original Sidewinder method, also adapted the approach to work from cheap, mass-produced DNA ingredients, further cutting time and cost.

  Wang and Robinson, together with Hie and entrepreneur Adrian Woolfson, cofounded a company called Genyro—to commercialize the technology, hoping to turn a profit through paying pharmaceutical and biotech clients. According to Robinson, however, they intend to make the Sidewinder platform broadly accessible to the academic research community.

  “We really want this to be an enabling platform,” says Robinson. “We want people to do cool things with the technology.”