AI Propels Rapid and Personalized Vaccine Development

Researchers rely on artificial intelligence to streamline research efforts and model viral protein structures. The data enables scientists to create personalized cancer vaccines and fast-track the development of a COVID-19 vaccine.

Image credit: DeepMind

Artificial intelligence has aided scientific researchers in their quest to develop vaccines for diseases such as cancer to viruses such as influenza and COVID-19. Machine learning technologies are helping scientists sift through the thousands of digital reams of research papers and studies that would otherwise delay progress for months, if not years. AI is also able to suggest components of a vaccine based on the viral protein structure, and computational modeling can predict how well a specific strain of virus or person will react to a potential vaccine.

It all starts with research and hypotheses, and AI has emerged as a critical tool on that front. Teams of researchers from the Allen Institute for AI and Google DeepMind have developed AI tools and shared data and research with scientists across the globe with the goal of advancing healthcare research. The Allen Institute for AI, founded by Paul G. Allen, created Semantic Scholar. Through the use of AI, it can scour the millions of pages of available research and pull together abstracts, tables, charts, and other relevant information far more efficiently than any human researcher possibly could.

To help COVID-19 researchers, the institute partnered with other organizations to create a database of more than 44,000 scholarly articles that relate to COVID-19 and other coronaviruses. The data is free and machine readable, so researchers can apply natural-language processing algorithms to make queries.

Google DeepMind has created AlphaFold, a deep learning system that uses AI and genetic sequencing to predict the 3D shape of a protein’s structure. Knowing the shape of a protein, an essential component of a virus, is critical to understanding how it works and, in turn, to developing drugs or vaccines to nullify the protein. In March, AlphaFold published its structure predictions for COVID-19 to help researchers in their quest to develop a vaccine. The system has had previous success: AlphaFold accurately predicted the spike protein structure of the SARS-CoV-2, the virus that caused the SARS outbreak.

Custom-made Drugs

Researchers at the University of Arizona Health Sciences are using AI to develop personalized cancer vaccines. By identifying and analyzing the specific mutations in a patient’s own cancer cells, the technology creates a set of genetic instructions that are added to a single molecule of messenger RNA (mRNA). The tailored mRNA is made into a vaccine that is customized for that specific person and that type of cancer. Initial tests indicate that the vaccines, when used in conjunction with immunotherapy, garnered a patient response rate of 50 percent, compared to 15 percent when immunotherapy alone is used.

The COVID-19 vaccines developed by Pfizer and Moderna also use mRNA technology. Early knowledge of the possible structure and shape of the protein helped fast-track vaccine development. Twenty-five different COVID-19 vaccines are under development, and Pfizer’s vaccine was recently approved for use in the UK and the US, less than one year after the emergence of COVID-19. Technology has helped propel the development of a vaccine that the world hopes will end the pandemic.

Predicting Vaccine Efficacy

The Pfizer and Moderna vaccines are a great start, and certainly better than no vaccine, but they may not be effective for all patients, depending on age, health, and even ethnicity. Researchers at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) used AI and machine learning to study a vaccine similar to those developed by Pfizer and Moderna. CSAIL found minority populations didn’t respond as strongly to the vaccine as white populations. The weaker response registered as less than 1 percent in white patients but reached nearly 10 percent in Asian populations. Using machine learning algorithms, the MIT researchers were able to predict where its vaccine might be less effective and to suggest possible strategies to improve the vaccine’s coverage across those populations.

The use of deep learning and computer modeling and analysis is changing how scientists can approach disease. In addition to speed, one of the greatest benefits of artificial intelligence to analyze scientific research is the ability to correlate data across those studies and identify potential angles, experiments, and treatments that might otherwise be overlooked. Coupled with structural models, AI has the potential to help eradicate a host of viruses and diseases.