Unleashing the power of artificial intelligence and machine learning for seasonal influenza

Influenza viruses remain one of the world’s greatest global health threats. The rapid evolution and spread of these viruses can lead to seasonal epidemics causing morbidity, mortality and a substantial socio-economic burden. Each year, an estimated 3–5 million severe cases and 290,000–650,000 respiratory fatalities worldwide are caused by influenza¹. GSK’s long-standing commitment to influenza prevention and innovation has driven advances in vaccine development. Today, GSK continues to harness its capabilities through innovative approaches to the development of new vaccines, such as artificial intelligence and machine learning (AI/ML), which are emerging as powerful tools to help combat these threats.

Navigating a rising threat with public-health implications

Young children, pregnant women, older adults and people with chronic comorbidities are particularly vulnerable to influenza-related complications, hospitalization and even death¹. By 2030, the World Health Organization expects the global population aged 60 years and older to increase by more than a third to 1.4 billion². While this trend of increased life expectancy can be considered as one of our most significant collective achievements, societal ageing also brings challenges. Indeed, as pointed out by the U.S. Centers for Disease Control and Prevention (CDC), immunity weakens with age, making older adults more susceptible to influenza and related complications; and those living with chronic noncommunicable diseases (such as diabetes and heart failure) or chronic respiratory illnesses are more vulnerable to infection, which may exacerbate underlying health conditions³.

The impacts of influenza infection have profound consequences on health systems and societies. For example, in the United States, influenza-related hospitalizations, outpatient visits and other interventions have created $11.2 billion in average annual healthcare costs⁴. Indirect costs, such as absenteeism and loss of productivity, pose other challenges to economies and the workforce. As populations age and the prevalence of noncommunicable diseases continues to rise, influenza may pose an even greater risk to health and could drive-up healthcare costs and workforce disruption. If left unaddressed, there may be serious consequences for public health and further stress on limited healthcare systems.

Annual vaccination is considered the best way to protect against influenza and its potentially serious complications, by significantly reducing the number of cases each year and helping to lessen the severity of the disease⁷. The Office of Health Economics (OHE) emphasized that adult immunization programmes provide a significant return on investment of up to 19 times their cost, as well as societal value⁵. These programmes help to reduce the costs associated with health complications for patients with co-morbidities, resulting in productivity gains and socio-economic advantages. Recent data from the OHE showed that countries that spent more on prevention and on immunization services reported lower mortality rates from vaccine-preventable diseases⁶.

Harnessing AI/ML for influenza vaccine development

Influenza epidemiology evolves every season. The viruses can mutate as they move across populations and geographies, and there is potential for circulating strains to differ from official vaccine recommendations.

Ensuring a precise alignment of the vaccine composition with the sequences of circulating strains is a key component to developing vaccines with better protection¹. The science driving research on new influenza vaccines can be improved by AI/ML⁷. Indeed, the rapid advancement of AI/ML tools is enabling scientists to analyze extensive multimodal datasets, extract critical insights and guide the development of influenza vaccines.

Integrating AI/ML technologies throughout the pipeline can propel development and improve existing vaccines⁷. These tools are currently supporting every development phase of GSK’s infectious disease drugs and vaccines, including those for influenza. As Pei-Yong Shi, vice president and global head of viral vaccines at GSK, pointed out, “These technologies have become critical for fast, timely and reliable data generation on viral strain evolution and prediction, and antigenic distance, to inform efficient decision-making in vaccine research and development (R&D).” For example, AI/ML tools could help enable a closer match between vaccine antigens and the strains projected to circulate in a given season, enhancing vaccine precision and the effectiveness of the induced immune responses. Incorporating genetic, structural, epidemiological and antigenicity data into deep-learning models can help identify emerging strains, predict their spread, and assess the antigenic distance of those in circulation. This approach also enhances the understanding of viral evolution.

Multiple AI/ML algorithms are needed to enable manufacturers to fast-track and de-risk their chemistry, manufacturing and controls (CMC) processes, including vaccine-quality testing methods. For example, in egg-based or messenger RNA (mRNA)–lipid nanoparticle (LNP) influenza vaccines, AI/ML tools may predict egg-passage adaptations of virus strains, enable rational design, or support rapid data analysis for quality assurance using long-read RNA nanopore sequencing^8,9,10. As mRNA CMC processes can be standardized across diverse vaccine products, this modality is particularly suited for digital twin modelling, which involves using a computerized replica of a real-world process. This type of modelling can be applied to help transition from batch to continuous in vitro mRNA transcription, boosting the end-to-end production of influenza vaccines.

AI has emerged as a transformative force in antigen selection and immunogen design. This is illustrated by AI/ML-based coding-sequence optimization that can enhance antigen design, stability and expression. Moreover, by employing deep mutational scanning of proteins, AI models can predict how antibodies bind to emerging variant strains. AI-powered models, such as generative adversarial networks, and molecular-dynamics simulations, can facilitate the rational design of novel influenza virus immunogens with optimized structural stability and antigen expression, and broader antigenic coverage¹¹. The impact of AI on antigen design has also been revolutionized by models such as AlphaFold. These approaches can predict three-dimensional (3D) models of viral proteins or peptides, including their folding and interactions with RNA, DNA and small molecules, allowing antigens to be strategically structured with optimized B- and T-cell epitopes.

To date, most investments in AI-driven R&D have concentrated on early discovery, leaving significant untapped potential in later-stage vaccine development. Indeed, multiple AI/ML applications—either in development or already established—have the potential to streamline clinical trials, making them faster and more cost-effective while producing higher-quality data⁴. As the innovations in AI/ML technologies continue to evolve swiftly, the role of these tools in later-stage processes and implementation is expected to grow exponentially.

A collaborative, AI-driven approach from disease surveillance to delivery

In addition to rigorous implementation, effective seasonal vaccination is also founded on high-quality epidemiological and health data, enabling a deep understanding of the burden of disease and peaks of influenza. In order to train AI/ML models to support seasonal influenza vaccination programmes, these vital data must be collected continuously to ensure relevant and timely outputs. Surveillance programmes that supply such datasets could be bolstered through additional data sources, such as human-mobility patterns or social media trends. As Shi noted, “Collaborations among stakeholders are particularly critical for surveillance and sharing of pathogen information, and multi-pronged approaches are needed to overcome existing challenges”. In fact, there are still many data gaps across different areas of vaccinology⁷. Vaccine manufacturers can help to close them by engaging in public–private partnerships such as the Human Immunome Project, which collects high-dimensional immune data from diverse populations to train AI algorithms in identifying genetic variants that affect vaccine responses¹² , and the Alliance for Genomic Discovery, which is dedicated to sequencing 250,000 DNA samples to enhance drug discovery and therapeutic development¹³.

Investing in AI-driven vaccinology may help to accelerate support solutions that match the pace of the dynamic challenges posed by influenza. By strengthening AI/ML-driven improvements in seasonal influenza data-collection, vaccine manufacturing and distribution, the benefits of existing programmes can be extended to more people and help reinforce scalable production and infrastructure on a global scale^1,4. New and sustained collaborations between vaccine manufacturers and multi-sector partners, such as academia, and engineering and AI/ML firms, can usher further innovations in coding sequence optimization and LNP design.

Strengthening trust in vaccines

Advances in AI/ML and technology platforms hold great promise for creating more-effective influenza vaccines. GSK is continuing to invest in cutting-edge technologies to design and develop vaccines with greater precision. Optimizing antigen design, manufacturing processes and other components of R&D are important approaches to improve vaccine development. Such capabilities equip us with more tools to address the medical needs of all patients, especially those who are the most vulnerable, such as older adults and individuals living with underlying medical conditions.

Innovation can also be a lever to reinvigorate public trust and confidence in vaccines—creating newer vaccines that can improve patient lives. When the positive impact of these efforts is felt every day, it bridges gaps between innovation and implementation.

Creating a collaborative and transparent environment for AI-driven vaccine development, in terms of datasets and models, can improve the influenza vaccine landscape. These technologies have heralded a new era in vaccinology, reshaping the development chain from pathogen surveillance through to vaccine delivery. Embracing the transformative potential of AI/ML offers a compelling pathway to protecting lives and supporting health systems that are impacted