Why is AlphaFold 3 such a hot topic in the pharmaceutical industry?

listen to article
5 minutes

This voice is automatically generated. Please let us know if you have one. feedback.

AlphaFold 3, an artificial intelligence program published by Google DeepMind and Isomorphic Labs earlier this month, isBreakthrough”, which has the potential to revolutionize drug discovery.

Patrick Bangert, senior vice president of data, analytics and AI at cloud services company Searce, says the hype is warranted.

“This is a Nobel Prize-worthy invention,” he said.

AlphaFold 3 features outlined in the journal Naturebuilds on two earlier AI models to help drug developers quickly identify promising targets, potentially cutting time to market by years.

Early versions of AlphaFold facilitated breakthroughs in drug development by making rapid and accurate predictions. 3D structure of proteinThis problem is a decades-old problem that has plagued computational biologists trying to understand how shape changes. affect protein function as an indicator of disease. Previous versions of AlphaFold are available through open source code. limited The main reason is that we cannot predict how proteins interact with other molecules throughout the cell. Still, they made great strides.

According to the magazine, “Millions of researchers around the world have used AlphaFold 2 to make discoveries in areas such as malaria vaccines, cancer treatments, and enzyme design.” Google AlphaFold 3 announcement. “AlphaFold has been cited more than 20,000 times and his scientific impact has been recognized through numerous awards, most recently Life Science Division Breakthrough Award”

The AlphaFold 3 program is AlphaFold server, It provides a web interface where non-programmers can enter the name of a protein or nucleic acid and generate a structural prediction containing important information. joint structure Use elements such as RNA, DNA, and ligands.

“For interactions between proteins and other molecule types, we have seen at least a 50% improvement compared to existing prediction methods, with prediction accuracy doubling for several important categories of interactions. ” said Google. However, accuracy could be further improved for other applications, such as protein-RNA interactions. Missing the point.

Traditionally, scientists have learned about protein structures using costly and time-consuming chemistry lab experiments, Bangert said. However, AlphaFold and competing programs, rose tafoldthe process is now faster and easier.

“Instead of physically trying out 100 options in a lab, you can try tens of thousands of options on a computer,” he said. “It still costs money, but it’s only a fraction of the cost, and of course it’s almost instantaneous.”

Where it normally takes companies 10 years to perfect a drug, this technology could do it in seven years, he said.

It’s hard to find fault with the AlphaFold 3, Bangert says.

“This is actually one of those rare cases where you can’t think of any drawbacks, mainly because this protein folding problem is so early in the drug development process,” he said. “This is before we do any chemical experiments, and definitely before any kind of clinical trials, and even if we make a mistake, it doesn’t matter because it doesn’t affect us in any way. So there are only positives. .”

That doesn’t mean companies using AlphaFold for drug development won’t face challenges.

“We need funding to make this happen. The next thing we need is data,” he said. “This is a protein fitting problem. You have a disease protein, then you have a candidate protein, and you try to fit them together. So to do that, you need enough data on the disease proteins available. .”

Doing so requires access to patient data, which creates regulatory hurdles, significant administrative burden and requires coordination with health systems, Bangert said.