AI Discovers Oral Drug That Blocks Multiple Coronaviruses

Scientists have developed a new oral antiviral drug using artificial intelligence and Hollywood animation software that can block infection from multiple coronaviruses, including SARS-CoV-2 variants, SARS, and MERS.

The compound, called WYS-694, reduced viral loads in infected mice by more than four-fold when given orally, offering hope for preventing future pandemic outbreaks. Unlike existing treatments that target easily mutated surface proteins, this drug zeroes in on hidden regions inside the virus that remain stable across coronavirus species.

The discovery emerged from an unusual collaboration between computational biologists, infectious disease experts, and drug developers at Harvard’s Wyss Institute who borrowed tools from the film industry to model how viruses hijack human cells.

Movie Magic Meets Viral Mechanics

Rather than studying static snapshots of viral proteins, the research team used Houdini—the same procedural animation software behind movie special effects—to create dynamic models of coronavirus spike proteins as they transform during infection. This approach revealed something crucial that previous methods had missed.

“We thought that constant regions that remain hidden while the virus initially binds to its host cell, but become accessible during a critical time window when it prepares itself for membrane fusion could be ideal sites,” explained Charles Reilly, the study’s first author and former Wyss Principal Scientist. “Targeting those could be a way to essentially lock the virus in at the pre-fusion stage before it can release its genetic material into the cytoplasm of host cells.”

The team’s molecular dynamics simulations showed that when a coronavirus spike protein binds to human cell receptors, mechanical tension causes a specific peptide sequence (TNFTISVTT) to physically separate from the virus’s S2 subunit. This separation exposes a hidden cavity that contains highly conserved residues—regions that don’t mutate because they’re essential for viral function.

Key Research Findings:

• WYS-694 is 12.5 times more potent than its predecessor compound
• Works against SARS-CoV-1, SARS-CoV-2, MERS, and multiple COVID variants
• Blocks viral entry independent of AXL kinase inhibition
• Demonstrates oral bioavailability with extended drug exposure
• Reduces viral load in infected mice by more than 4-fold

From Screen to Lab Bench

Using this mechanistic insight, researchers computationally screened about 10,000 existing drugs to find molecules that could bind tightly to the newly identified pocket. The highest-ranking oral drug was bemcentinib, an FDA-approved cancer medication that had previously shown anti-COVID activity in clinical trials.

However, bemcentinib works by inhibiting AXL kinase, making it unclear whether its antiviral effects came from this known mechanism or from binding the viral protein directly. To settle this question, medicinal chemist Joel Moore designed structurally similar compounds that retained viral binding but lost AXL activity.

The resulting molecule, WYS-633, proved that the antiviral activity operated independently of kinase inhibition. “To unequivocally prove that bemcentinib achieved its antiviral activity through our proposed mechanism and further improve its efficacy, we needed to create structurally similar compounds (analogs) that lack any affinity to AXL but retain their affinity to the Spike protein’s binding pocket,” said Reilly.

Mechanical Forces Drive Viral Entry

What makes this research particularly novel is its focus on the mechanical forces involved in viral infection. The team’s simulations suggest that when coronavirus spike proteins bind to human cell receptors, the resulting tension triggers conformational changes necessary for membrane fusion.

This mechanical unfolding process appears critical across multiple virus families. The researchers noted that similar force-dependent mechanisms operate in influenza, HIV, and other envelope viruses, suggesting their approach could have broader applications.

Through advanced steered molecular dynamics simulations, they demonstrated that WYS-694 stabilizes key structural elements that would otherwise rearrange to form the virus’s post-fusion state. The drug essentially acts as a molecular “wrench” that prevents the spike protein from completing its shape-shifting sequence.

Beyond COVID: Broad-Spectrum Protection

Testing revealed that WYS-694 blocked infection by multiple coronavirus variants of concern, including Alpha, Beta, Delta, Gamma, and Omicron strains. The compound also prevented entry by SARS-CoV-1 and MERS-CoV, demonstrating true broad-spectrum activity.

This versatility stems from targeting conserved internal regions rather than the variable surface domains that current antibody therapies recognize. Since these hidden regions don’t face immune pressure, they mutate far less frequently than exposed binding sites.

The pharmacokinetic profile of WYS-694 also showed promise for practical use. When administered orally to mice, the drug achieved extended exposure with a time to peak concentration of 24 hours, potentially allowing for convenient once-daily dosing.

Validation Through Advanced AI

To further validate their findings, the researchers employed Google’s AlphaFold 3 machine learning algorithm to predict protein structures in the presence of their drug compound. These predictions consistently positioned WYS-694 within the target region when the S1 subunit was displaced, supporting their mechanistic hypothesis.

Interestingly, when AlphaFold 3 modeled the intact spike protein with the S1 region present, it did not place the drug in the target site, confirming that mechanical displacement of S1 exposes the binding pocket. This computational validation aligned perfectly with their experimental observations.

From Defense Funding to Pandemic Preparedness

The project began in spring 2020 with emergency support from the Defense Advanced Research Projects Agency (DARPA) as the COVID-19 pandemic unfolded. The urgency of the crisis drove the team to focus initially on repurposing existing FDA-approved drugs rather than designing entirely new molecules.

This rapid-response approach led to the identification of bemcentinib within months, demonstrating how integrated computational and experimental pipelines can accelerate drug discovery during health emergencies.

Why might this mechanical approach succeed where other strategies have struggled? Traditional antiviral development typically targets static protein structures, but viruses are inherently dynamic machines. By focusing on the mechanical transformations required for infection, the researchers identified intervention points that remain hidden from conventional screening methods.

Pipeline for Future Outbreaks

The methodology developed for this project extends beyond coronaviruses. The same integrated approach—combining molecular dynamics, AI-based docking, evolutionary analysis, and medicinal chemistry—could potentially target other viruses that rely on membrane fusion for entry.

“This approach holds great potential for the discovery of drugs against a number of other virus families utilizing membrane fusion proteins, including influenza, HIV, Ebola, Measles, and others,” noted senior author Donald Ingber, who directs the Wyss Institute.

The researchers designed their computational pipeline to be modular and adaptable, capable of incorporating new AI techniques as they emerge. This flexibility could prove crucial for responding to future viral threats that may require rapid countermeasure development.

The Road Ahead

While WYS-694 shows impressive preclinical results, important questions remain. The researchers acknowledge they haven’t yet demonstrated direct binding to the predicted target site through high-resolution structural studies. Such confirmation would require techniques like cryo-electron microscopy or hydrogen exchange mass spectrometry.

The team also noted limitations in their current computational approach, which prioritized rapid hypothesis generation over detailed quantitative analysis. Future studies could employ more sophisticated simulation methods to better understand the precise energetics of drug binding and viral transformation.

Nevertheless, the current results provide compelling evidence for a new class of antiviral compounds that target mechanical aspects of viral entry. As the research team noted, this represents “the first in a new class of antiviral compounds that target mechanical transformations within viral S proteins required for membrane fusion and cell entry.”

For a world still grappling with COVID-19 and bracing for future pandemic threats, the development of orally available drugs with broad-spectrum coronavirus activity offers a valuable addition to our antiviral arsenal. The ability to prevent infection prophylactically, particularly in regions where vaccine access remains limited, could prove crucial for pandemic preparedness.

The integration of AI, physics-based modeling, and traditional drug development showcases how interdisciplinary approaches can tackle complex biological challenges. By borrowing tools from Hollywood and combining them with cutting-edge computational biology, these researchers have opened new pathways for antiviral discovery that could reshape how we prepare for future infectious disease outbreaks.