The Human Vaccines Project combines biology with artificial intelligence (AI) to understand the human immune system. How do you guarantee safety when using AI in human health?
Wayne Koff: COVID-19 has painfully demonstrated the vulnerability of ageing and of low and middle-income countries’ (LMIC) populations to disease, the global inequities in tackling diseases in LMICs, and how unprepared the world is when it comes to preventing pandemics. Scientists have been warning us about these issues for decades. With every vaccine, there remains the very real challenge of protecting those who are most vulnerable to disease: newborns, ageing adults and people living in LMICs.
Most vaccines are not as effective for these vulnerable groups, and the track record for vaccine uptake is historically low for both ageing and LMIC populations. This problem is not unique to COVID-19. From infectious diseases to chronic conditions like cancer, heart disease and Alzheimer’s, the world is facing a staggering increase in disease burden over the coming decades, as the median age of the global population is increasing dramatically.
To protect the world’s increasingly vulnerable society against the diseases of the 21st century, we at the Human Vaccines Project are developing new strategies to develop and deploy vaccines. This strategy is based on the fundamental principle that deciphering effective immunity in vulnerable population groups is central to improving health in these populations, that decoding the human immune system is now feasible with the convergence of biomedical, engineering and computer sciences (AI and machine learning), and that global coordination will facilitate achieving these objectives. The current pandemic has highlighted how urgent this mission is.
The amount of data generated in both scientific literature and scientific studies now exceeds human ability to fully understand and analyse such data. AI can help analyse, draw conclusions and generate new research hypotheses from increasingly massive datasets, and can help speed up the process of running scientific studies and generating conclusions from them, including new ways of developing safe and effective vaccines against deadly diseases.
We can think of AI as a tool to understand complex data brought about by advances in the fields of mathematics and statistics, one of many that scientists use to understand biological systems. Despite the great potential for AI models of human immunity to transform product development, making it faster, less expensive and with a greater probably of success, in the immediate future we will still need to rely on well-established, ethical and safe processes to test and approve new drugs and vaccines.
What has been the Human Vaccines Project’s most challenging and rewarding discovery in new diagnostics?
Wayne Koff: By better understanding how the human immune system fights disease, we can revolutionise the development of new vaccines and therapies, and ultimately our approach to human health. By understanding our own immunity, we have the potential to better protect ourselves from the next pandemic, while making advances against intractable diseases like cancer, HIV and Alzheimer’s. The work is hard and complex, but the rewards are immense.
The most rewarding discoveries thus far in the field of immunomics championed by the Human Vaccines Human Immunomics Initiative – Human Vaccines Project (where we harness the convergence of human immunology, systems biology, AI and machine learning) are the following:
1) The immune receptor repertoire (i.e. the universe of primary cellular receptors in the human immune system), is more common across individuals than originally thought. This improves the scope for vaccine efficacy, making them not only work for subsets of the population (e.g. 50 % effective vaccines), but for the whole population (e.g. 90 %-100 % effective vaccines);
2) Healthy immune systems can be differentiated from diseased ones (i.e. those affected by autoimmune diseases), enabling the development of a new field of immune repertoire-based diagnostics;
3) Baseline immunomic setpoints (biomarkers) predict the outcome of vaccine-induced responses, and, by extension, how we respond to pathogens – highlighted by the current COVID-19 pandemic. This enables us to predict who will respond better to immunisation prior to vaccination and therefore open the door to immune modulation so as to enhance vaccine immunity for all.
Does the COVID Vaccine Initiative show any promising results in fighting the pandemic? Why does it focus on vulnerable populations?
Wayne Koff: The development of numerous successful vaccines against COVID-19 is not just a remarkable scientific achievement: it also gives us a chance to understand why some vaccines are so successful and others are not. This will help us design even better vaccines in the future to help fight new pandemics and old diseases, as well as protect those who are most vulnerable to disease.
We know from many generations of vaccine research that not everyone responds equally well to vaccines because of differences in our immune system. For example, we have not been very successful at protecting newborns from infection because their immune system is still immature. We also know that older adults and the elderly often have poorer responses to vaccines because their immune system is ageing. If we could figure out how to better protect these vulnerable populations, we could save millions of lives. We published a comprehensive review on this topic in Science Translational Medicine.
How can we avoid a future pandemic caused by a respiratory virus with the help of AI and coding?
Wayne Koff: AI can help us design better, safer and faster vaccines. We can do this by running thousands of simulations or designs on a computer in a matter of months, identifying the ones with the best chance of success, and beginning testing under the standard product development process that worked so well in the COVID-19 pandemic. This can help the process of vaccine development move even faster and allow for a greater chance of success. As we learned with COVID-19, we could have prevented hundreds of thousands of deaths if we had had a vaccine a few months earlier. Thinking even further ahead, we have had three major global coronavirus outbreaks in the last 20 years (SARS-1, MERS, and SARS-2/COVID), and we are worried that more could be coming. We need the vaccines before the next outbreak occurs. We believe that AI could be a powerful tool to help design universal vaccines that work across all coronaviruses threats, past and future. This is the real power of AI: it can accelerate science and ultimately save lives.
What can be done to close the gap between political and public health figures, so that there is transparency, education and commitment to ensuring greater vaccine confidence?
Wayne Koff: We have been dealing with narratives when it comes to vaccines. On the one hand, immunisation is considered one of the top ten public health achievements of the 20th century. On the other hand, vaccine hesitancy was among the ten threats to global health in 2019 even prior to the COVID-19 pandemic. There was already a big geographic variation in terms of vaccine confidence in the pre-COVID-19 era.
How can you explain this?
Wayne Koff: To understand this phenomenon, we have to understand that vaccine hesitancy is not new; it has been around since vaccines were introduced, and can be explained by the fact that vaccines are products that are administered to healthy people. Vaccine hesitancy is a multifaceted phenomenon that stems from fears and doubts that people may have as regards the risk-benefit analysis of vaccines, in particular in terms of their effectiveness and safety. We see this in particular with new vaccines. A lack of knowledge and awareness, misperceptions of personal risk and the belief that the threat of disease is exaggerated can all affect how people perceive disease. Risk perception of both disease threats and vaccines have always been embedded in historical and socio-cultural contexts.
Furthermore, vaccine confidence reflects a set of relationships with individuals as part of a community decision. When thinking about vaccine hesitancy, it is important to differentiate between misinformation and mistrust. Even a decade after the retraction of a faulty paper that erroneously linked autism with the measles, mumps, rubella vaccine, the misconception that vaccines cause autism still exists.
The internet and social media changed the way people consumed this type of information and gave rise to the anti-vaccine communities that exist today, largely in rich countries. In some other countries, people mistrust vaccines due to religious, political, and historical factors. There are many different elements contributing to these beliefs and perceptions.
Is COVID-19 vaccine acceptance any different?
Wayne Koff: The scepticism around this particular vaccine goes well beyond typical vaccine hesitancy. With COVID-19, we are dealing with a new disease and a lot of uncertainty, not to mention the fact that innovation and vaccine discovery are happening at an unprecedented speed. That is a lot for the public to process, especially with the misinformation that exists. Our research has found that COVID-19 vaccine acceptance varies considerably around the world. Key attributes have shaped differential profiles in different subgroups. Doubts surrounding the disease, scepticism of a new vaccine’s safety and effectiveness, distrust of the system along with previously shaped vaccination beliefs prior to COVID-19 have interplayed and together contributed to the reluctance we are observing today. These key attributes apply to all countries in our study, as well as across major US subgroups.
In general, COVID-19 vaccine acceptance is higher in many of the middle-income countries, in comparison to rich countries. In the USA and Europe, the public has become much less familiar with the horrors of infectious diseases until recently. A high burden of disease alone may not provide sufficient motivation for the public to seek vaccination for themselves, especially if there has been COVID-19 denial due to political reasons.
Vaccine acceptance also depends on how the government handled or mishandled the pandemic, on whether there is widespread public trust of the national government and, more broadly speaking, on the country’s public health culture. It has a lot to do with the approach to public health, as we have seen with mask wearing and other compliance issues. It is not surprising that people who are not complying with public health measures are less willing to accept a vaccine.
Marginalised subgroups such as low income, low-educated, non-insured and minority groups are generally more vulnerable to COVID-19 vaccine hesitancy. In the USA, women, young people, Republicans, African Americans and Latino Americans tend to be more reluctant than the rest of the population. Studies have also found pockets of highly educated social media subgroups with a high level of reluctance in India, among other countries. Social media and social networks have played a key role in shaping perceptions.
Because of the enormous amount of information on COVID-19, people are becoming more knowledgeable about vaccination terminology: herd immunity, partial efficacy, side effects, halting trials, etc. On the other hand, there is no coherent and comprehensive base knowledge for lay audiences. This lack of foundational knowledge combined with the fear and anxiety caused by the pandemic has fuelled the spread of misinformation.
How do you address COVID-19 vaccine hesitancy?
Wayne Koff: First of all, our findings suggest that there is room for governments, especially those in countries currently experiencing COVID-19 denial, to correct the perception of the disease in the coming months as vaccine rollout continues. Policy makers must address denial and doubt of the disease and public fears and misconceptions by using clear and unified communication. This will result in the national consensus on the utter importance of public health measures, including the wearing of facemasks and mass vaccination to end the crisis caused by the pandemic. The public’s risk perception and confidence in a new vaccine can therefore be put into perspective, making it easier to see what is truly at stake. Vaccine education campaigns should focus on the pandemic as a whole and its impact on communities, rather than limiting their content to vaccine safety and effectiveness.
Secondly, it is essential to (re-)establish trust in science, institutions and government messages and messengers. The dynamics of trust are fundamental to effective policy response. This requires transparency from public authorities on what basic policies are being developed and how they are implemented. Another crucial element is having the courage to say that some measures were not suitable or that they were insufficient.
Thirdly, due to geographic variation in vaccine acceptance, we need to develop vaccine programmes that are specific to each country and their regions. Effective community engagement is essential. Lessons learned from previous new vaccine rollouts, including the human papillomavirus (HPV) vaccine rollout and the Ebola vaccine trial have all underscored the importance of community work that takes into account the sensitivities of specific countries, regions and subgroups.
Fourth, because of the need to truly understand the context of each community, their perceptions and peer dynamics surrounding vaccination, it is important to invest in research in this field and develop more effective research-based communication strategies. It is particularly important to work with social media platforms and influencers when conducting research on online communities and strategies.
Fifth, significant portions of pregnant women and mothers have expressed additional safety concerns due to insufficient pregnancy and child-specific clinical evidence. As more such data become available for these two groups, there will be more opportunities for pregnant women and mothers to build trust on the scientific approval of these vaccines. Vaccination campaigns for women and children should be specific to each country in order to achieve maximum impact.