A California company has used artificial intelligence to identify a forgotten breast cancer compound as a potential broad-spectrum antiviral that works against the viruses causing colds, flu, COVID-19, RSV, norovirus, and several forms of hepatitis. The drug, now called MDL-001, was originally developed as ERA-923 in the early 2000s but abandoned after disappointing results as a selective estrogen receptor modulator. Model Medicines' AI platform mined decades of chemical, biological, and clinical data to predict that the molecule could bind to the Thumb-1 domain of RNA-dependent RNA polymerase (RdRp), an enzyme many viruses use to copy their genomes.

The preclinical work, which will be presented at the European Congress of Clinical Microbiology and Infectious Diseases in April, is not yet peer-reviewed. Methodology involved testing the compound in infected cell cultures against multiple viral families and in mouse models of COVID-19 and hepatitis. In mice, MDL-001 reduced viral loads in lungs and livers and lessened weight loss, though exact sample sizes and statistical details were not disclosed in the New Scientist coverage. This lack of transparency is a limitation typical of early company-reported data.

What the original reporting under-emphasized is the long history of failure in this exact approach. Previous Thumb-1 inhibitors developed for hepatitis C showed little activity against other viruses, as noted by University of New South Wales virologist Peter White. Expert Daniel Rawle correctly flagged that most antivirals effective in vitro fail when moved into living organisms. The coverage also missed the broader context: while remdesivir targets RdRp, it requires intravenous delivery and has narrow approved uses; molnupiravir is an oral COVID drug but carries mutation risks and is not broadly effective.

Synthesizing this with a 2021 Nature Reviews Drug Discovery article on antiviral drug development and a 2019 review in Antiviral Research on polymerase inhibitors reveals a consistent pattern. Conserved viral targets are attractive but notoriously difficult to hit without host toxicity or rapid resistance. Model Medicines claims MDL-001 has a unique docking mode enabling pan-viral activity, yet this remains to be independently verified.

The editorial lens here is clear: a single safe, oral broad-spectrum antiviral would transform public health by removing the need for diagnostics, reducing unnecessary antibiotic prescriptions for viral illnesses, keeping people productive during seasonal outbreaks, and providing a rapid-response tool for future pandemics. It would simplify treatment the way broad-spectrum antibiotics did for bacterial infections. However, this same breadth raises legitimate concerns about widespread use driving resistance across multiple virus families and potential off-target effects on human RNA processing.

The AI-driven repurposing itself is noteworthy. By training on vast public datasets of failed compounds, the platform turned a clinical dead-end into a candidate with minimal side effects from prior human cancer trials. Yet AI predictions still require rigorous validation; many similar 'breakthroughs' have faded in phase 2 and 3 trials. If MDL-001 succeeds in the safety trial planned for early next year and subsequent efficacy studies, it could mark a genuine shift. Until then, healthy skepticism is warranted. The gap between lab promise and real-world impact remains wide, and history suggests cautious optimism is the responsible stance.