This article is more than 1 year old

How Pfizer used AI and supercomputers to design COVID-19 vaccine, tablet

Everything from molecular structure to freezer distribution involved ML, says tech chief

GTC AI algorithms were involved in every step of the way to design and deliver Pfizer's COVID-19 vaccine and antiviral tablets to hundreds of millions of people as coronavirus spread around the world.

Lidia Fonseca, Pfizer's chief digital and technology officer, said machine-learning technology was crucial in speeding up the pharma-goliath's processes to develop the vaccine: starting from drug discovery, to clinical trials, to supply chain management, to, finally, distribution to people.

Pfizer developed two weapons to fight COVID-19: a vaccine developed together with BioNTech, sold under the name Comirnaty, and the less common oral antiviral treatment, Paxlovid.

"We have clear examples in bringing both the vaccine, Comirnaty, and now the oral treatment, Paxlovid, to COVID breakthrough treatments to market in record time by harnessing the power of digital data in AI," she said in a conversation with Joe Ucuzoglu, chief executive officer at Deloitte US, on Tuesday at Nvidia's GPU Technology Conference.

Machine-learning models are well-suited for the task of finding novel molecules with a required set of properties. Scientists were able to sequence and piece together the structure of the coronavirus' spike protein early on in the pandemic. These spikes latch onto the surface of cells, where the coronavirus is absorbed and allowed to replicate. The host human then becomes infected, causing respiratory illness, fatigue, headaches, and other detrimental health effects.

An mRNA vaccine, like Pfizer's, works by teaching our cells to make just the spike protein part of the virus so that your immune system can learn to destroy such spikes on sight in future, thus hopefully eliminating the coronavirus should it enter the body.

Fonseca said Pfizer was able to design its mRNA-based vaccine and get it into clinical trials in just four months. Machine-learning algorithms helped the company predict yields during the manufacturing stage before tens of thousands of volunteers from six countries were recruited for testing. AI systems were used to analyse any discrepancies in the participants' symptoms.

"Additionally, we use both AI and machine learning to predict product temperatures, and enable preventative maintenance for the more than 3,000 freezers that house our vaccine doses. And we also leverage IoT and sensors to monitor and track vaccine shipments and temperatures at close to 100 per cent accuracy," she added.

By the time Pfizer turned its attention to making antiviral Paxlovid – which can be taken when symptoms begin to lessen the blow of the virus – scientists had a better idea of how to minimize allergic reactions observed in some people who had been injected with its vaccine. After their models generated a range of promising candidate molecules, they were able to test a fraction of them in virtual simulations run on supercomputers. 

"Many of the allergic reactions that clinical trial participants reported while testing our vaccine resulted from certain certain liquid lipid nanoparticles in the vaccine itself, using supercomputing we ran molecular dynamics simulations to find the right combination of lipid nanoparticle properties that reduce allergic reactions," Fonseca confirmed.

Pfizer reportedly turned to the MareNostrum 4 supercomputer to test potential new drugs virtually, according to HPC Wire. The big system is housed at the Barcelona Supercomputing Center, and has a ​​peak performance of 11.15 Petaflops; it contains 3,456 nodes each made up of two Intel Xeon Platinum 8160 processors and four Nvidia V100 GPUs.

Big Pharma is increasingly turning to AI. The technology automates processes over large scales, cutting costs, and reducing the time to get new drugs to market. Large manufacturers like Pfizer will have to work with smaller outfits specializing in areas the bigger businesses are weak in, such as data generation, aggregation, or analytics, Fonseca said.

"AI and machine learning powered data analytics are turbocharging drug discovery and development, but they can also help enhance prevention, early detection, personalized treatment and digital therapies. On the technology front, developments such as new gene therapies, digital therapeutics [will be] approved and reimbursable. Quantum computing capabilities will help bring breakthrough medicines to patients faster."

"I believe COVID-19 has accelerated these trends by as much as five years," she added. ®

More about


Send us news

Other stories you might like