AI predicts vaccines in real time, but can it be trusted?

Written by By Rosa Inocencio Smith, CNN Computational models predict the effectiveness of vaccines and drug treatments for several diseases. One of the most impressive models ever developed is the MRV Vaccine from National…

AI predicts vaccines in real time, but can it be trusted?

Written by By Rosa Inocencio Smith, CNN

Computational models predict the effectiveness of vaccines and drug treatments for several diseases. One of the most impressive models ever developed is the MRV Vaccine from National Cancer Institute (NCI). The project predicts the effects of several immunotherapies — including chemotherapy, radiation therapy and thimerosal — for more than 300 cancer types.

The health care industry is constantly fighting for further improvement, and now the team at Moderna have developed an AI-powered model of their own. It predicts the efficacy of immunotherapies such as vaccines, only by working together with the person the vaccine is intended to protect — rather than against the illness the vaccine is meant to protect against.

The system can be used to evaluate the effectiveness of hundreds of immunotherapies. It uses a battery of machine learning algorithms in combination with public data and AI to predict a person’s response to different drugs.

This system, dubbed Agile Vaccine Environments, uses models that are trained to predict how to improve vaccination — not just its effectiveness, but also its elimination from the body. The system was tested on 35 types of vaccine and immunotherapies.

The first study was published in the journal Science. The second study, funded by the National Institute of Health, will be published in 2020.

The results show that the Vaccine Protection (PE) loss rate was 36%, which is double that of the previous CR0 models that showed a two-to-three percent reduction in PI loss (loss of protection). There were no samples left for review with the new model, making it hard to say exactly why the difference was seen. However, the authors do report that the model had a greater success rate in analyzing immunotherapy than previously, with a corresponding decrease in vaccine loss.

The new Agile Vaccine Environments is still in its pilot stage and subject to a rigorous testing cycle that entails simulated exposure to drugs, and also real-life clinical trials. The use of Agile Vaccine Environments for vaccination-related research will be reviewed in 2020 and evaluated further in 2022.

Experts stress that CR0s are the “prime movers” for new models, in terms of new advances in vaccine safety. In addition, the CR0s suggest newer vaccines and drug candidates, allowing for improved vaccine formulations, with better adherence to the vaccination schedule. However, for Agile Vaccine Environments, the use of several-phased models enables researchers to generate predictions for more than 600 different different types of vaccines.

Leave a Comment