Skip to Content
MIT Technology Review

null
  • Joo Chuan Tong

    Age:
    31

    In recent years, Asia has been the epicenter of many emerging and reëmerging diseases, including avian influenza, severe acute respiratory syndrome (SARS), malaria, and chikungunya. The 2003 SARS epidemic, which coincided with the start of my PhD at the National University of Singapore’s department of biochemistry, left a particular impression on me. Thus began my quest for more effective ways to create vaccines to combat such diseases.

    Credit: Courtesy Joo Chuan Tong

    Vaccination is a powerful tool. Each person’s immune system is unique, however, and vaccines do not take these individual differences into account. So although today’s vaccines protect the majority, some people fail to develop immunity, while others may have adverse reactions. At the same time, rapidly mutating bacteria and viruses evolve to evade immune protection. Every time a new strain emerges, a new vaccine must be created, as with the annual flu shot.

    If we could map out the genetic profile of each individual’s immune system, efficiently create vaccines against the newest strains of a disease, and match the two, we would stand a far better chance of protecting people. At the Institute for Infocomm Research, I lead a team that is developing computer algorithms to help make this dream of personalized vaccines possible.

    Our bodies rely on proteins called human leukocyte antigens (HLAs) to recognize foreign substances (i.e., antigens) from disease-­causing microbes and marshal our immune systems against them. These same proteins process the antigens in vaccines, triggering resistance. But there are thousands of variants of the 11 HLA proteins, and each person inherits at most two of the possible variants for each one. Our algorithms take those genetic differences into account to help select antigens that are most effective in triggering an immune response.

    We begin by creating 3-D models of the interactions between different HLA molecules and antigens. We then use those models to train machine-learning algorithms to identify antigens likely to bind to the largest variety of HLA molecules; those antigens have the best potential to be effective vaccines. Our goal is to create models of the 120 to 150 most common HLA variants, which should cover 95 percent of the global population. By matching possible antigens with the HLA variants most common in a population, vaccines may be tailored to specific groups or, with personal screening, even to individuals. –Joo Chuan Tong