AI-designed nasal antiviral platform targets broad protection against COVID-19, flu, other respiratory viruses

KAIST team uses AI-driven protein design and advanced intranasal delivery to stabilise interferon-lambda to achieve a prolonged level of mucosal retention with strong antiviral effects in preclinical models

Respiratory viruses that display extensive strain diversity and rapid mutation – including influenza and SARS‑CoV‑2 – have continued to present challenges for vaccine-based control. Antigenic drift and shift can erode vaccine effectiveness, while global deployment often depends on cold-chain infrastructure that can be unreliable in some regions. Against this backdrop, researchers at the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea, have reported the development of a nasal antiviral platform designed to provide broad, strain-agnostic mucosal protection at the earliest stage of infection.

A joint research team, led by Professor Ho Min Kim and Professor Hyun Jung Chung from the Department of Biological Sciences, together with Professor Ji Eun Oh from the Graduate School of Medical Science and Engineering, has used artificial intelligence (AI) based protein design to re-engineer interferon-lambda for intranasal use. The redesigned protein has then been combined with a delivery system intended to enable rapid diffusion and prolonged retention within nasal mucosa. The team has described the approach as a universal preventative platform against diverse respiratory viruses.

Interferon-lambda is an innate immune signalling protein produced naturally by the body to suppress viral replication, particularly at epithelial barriers such as those lining the respiratory tract. It has played an important role in host defence against pathogens responsible for the common cold, influenza and COVID-2019. However, despite its biological relevance, prior attempts to formulate interferon-lambda as a nasal therapeutic have been limited by poor stability and short residence time. Heat sensitivity, enzymatic degradation, mucus entrapment and clearance by ciliary motion have all reduced efficacy in practice.

To address these constraints, the KAIST team has applied AI-driven protein design to reinforce the structural vulnerabilities of interferon-lambda. Using computational modelling, the researchers identified flexible loop regions within the protein that were prone to instability. These regions were redesigned into more rigid alpha-helical structures, which stabilised the overall fold.

The redesign also addressed protein aggregation, a common issue that can reduce bioactivity and complicate formulation. Through surface engineering, the researchers altered the protein exterior to improve its compatibility with water and reduce the tendency for individual molecules to adhere to one another. In parallel, glycoengineering was used by adding specific glycan structures to the protein surface to further enhance robustness and resistance to degradation.

According to the team, the resulting interferon-lambda variant has shown a marked improvement in physicochemical stability. In laboratory testing, the engineered protein remained intact for up to two weeks at 50 degrees Celsius and retained the ability to diffuse rapidly through viscous nasal mucus, a critical requirement for effective intranasal delivery.

To extend retention within the nasal cavity, the researchers encapsulated the engineered interferon-lambda within nanoliposomes and coated the particles with low molecular weight chitosan – which is a sugar that comes from the outer skeleton of shellfish, including crab, lobster, and shrimp. This combination substantially increased mucoadhesion, enabling the formulation to adhere to the nasal epithelium for an extended period rather than being cleared rapidly.

In animal models that had been infected with influenza virus, the application of the intranasal platform produced a strong antiviral effect. Viral levels within the nasal cavity fell by more than 85 per cent compared with the untreated controls, therefore showing strong effectiveness to suppress infection at its point of entry.

The researchers have characterised the platform as a mucosal immune strategy designed to intercept viral infection at an early stage through simple nasal administration. They have suggested that such an approach could complement existing vaccines by providing rapid, non-specific protection against both seasonal respiratory viruses and emergent or mutated strains for which vaccines may not yet be available.

“Through AI-based protein design and mucosal delivery technology, we simultaneously overcame the stability and retention time limitations of existing interferon-lambda treatments,” Professor Ho Min Kim said.

“This platform, which is stable at high temperatures and remains in the mucosa for a long time, represents an innovative technology that could be used even in developing countries that lack strict cold-chain infrastructure. It also offers significant scalability for the development of a range of therapeutics and vaccines,” he added.

The team further said that the work reflected: “a meaningful achievement resulting from multidisciplinary convergence research, spanning AI protein design, drug delivery optimisation and immune evaluation in infection models”.

The research involved Dr Jeongwon Yun from the KAIST InnoCORE institute, formally titled the AI Co-Research and Education for Innovative Drug Institute, Dr Seungju Yang from the Department of Biological Sciences, and doctoral researcher Jae Hyuk Kwon from the Graduate School of Medical Science and Engineering, all of whom contributed as co-first authors.

For further reading please visit: 10.1002/advs.202506764