Meta AI Unlocks Hundreds of Millions of Proteins To Aid Drug Discovery (wsj.com) 11
Facebook parent company Meta Platforms has created a tool to predict the structure of hundreds of millions of proteins using artificial intelligence. Researchers say it promises to deepen scientists' understanding of biology, and perhaps speed the discovery of new drugs. From a report: Meta's research arm, Meta AI, used the new AI-based computer program known as ESMFold to create a public database of 617 million predicted proteins. Proteins are the building blocks of life and of many medicines, required for the function of tissues, organs and cells. Drugs based on proteins are used to treat heart disease, certain cancers and HIV, among other illnesses, and many pharmaceutical companies have begun to pursue new drugs with artificial intelligence. Using AI to predict protein structures is expected to not only boost the effectiveness of existing drugs and drug candidates but also help discover molecules that could treat diseases whose cures have remained elusive.
With ESMFold, Meta is squaring off against another protein-prediction computer model known as AlphaFold from DeepMind Technologies, a subsidiary of Google parent Alphabet. AlphaFold said last year that its database has 214 million predicted proteins that could help accelerate drug discovery. Meta says ESMFold is 60 times faster than AlphaFold, but less accurate. The ESMFold database is larger because it made predictions from genetic sequences that hadn't been studied previously. Predicting a protein's structure can help scientists understand its biological function, according to Alexander Rives, co-author of a study published Thursday in the journal Science and a research scientist at Meta AI. Meta had previously released the paper describing ESMFold in November 2022 on a preprint server. Further reading: What metaverse? Meta says its single largest investment is now in 'advancing AI.'
With ESMFold, Meta is squaring off against another protein-prediction computer model known as AlphaFold from DeepMind Technologies, a subsidiary of Google parent Alphabet. AlphaFold said last year that its database has 214 million predicted proteins that could help accelerate drug discovery. Meta says ESMFold is 60 times faster than AlphaFold, but less accurate. The ESMFold database is larger because it made predictions from genetic sequences that hadn't been studied previously. Predicting a protein's structure can help scientists understand its biological function, according to Alexander Rives, co-author of a study published Thursday in the journal Science and a research scientist at Meta AI. Meta had previously released the paper describing ESMFold in November 2022 on a preprint server. Further reading: What metaverse? Meta says its single largest investment is now in 'advancing AI.'
Side effects (Score:2)
For most debieses it seems we have a lot of different drugs, the drawback being side effects are death by cancer, death by heart failure, and death by dehydration. If the AI can determine which proteins will provide results without the side effects that would be very good.
Re: (Score:2)
If the AI can determine which proteins will provide results without the side effects that would be very good.
The problem is that "human" isn't an exact target. Despite our similarities, we have differences that cause us to react to drugs differently. I think it might be more helpful to approach the problem in reverse and classify DNA to identify the common factors that result in differing reactions.
I would be very interested in learning why I have a seizure type response to antihistamines.
Comment removed (Score:4, Interesting)
Re: (Score:2)
I mean that in a positive way. Grey goo and nanorobots are an interesting concept in science fiction, but the simple fact is the level of mechanical and chemical complexity in proteins, ribosomes, enzymes, viruses and the likes dwarfs even the most fantastical ideas that nanomachines could even hope to achieve.
Take CRISPR. CRISPR's function is to sift through an entire genome, find a target sequence and destroy it, and it can unwind the entire
Re: (Score:2)
When you get small enough the difference between bioengineering and nanotechnology is null.
The difference is that with bioengineering, you have to settle for one of the ways that nature arranges atoms, e.g. protein folding.
Premature headline (Score:3)
Seems like a severely premature headline to me.
Sure, maybe they predicted how hundreds of millions of proteins will fold... but how many have they actually confirmed?
Predictions are cheap. In any complex system even those born of genuine understanding and long experience are usually wrong.
Re: (Score:2)
Re: Premature headline (Score:1)
Re: (Score:3)
They checked the ability to predict the conformation of proteins by comparing with known protein structures. As in, there's a big list of proteins we know the structure of, so they confirmed that it can predict those shapes correctly. Sure there may be proteins it can royally F up on but if it got that list right, it seems unlikely to screw up on most other proteins.
It's like if you tell me you know the lyrics to every song in the world. If I ask you the few that I randomly happen to know, I would assume yo
Re: (Score:2)
So, what I'm hearing is that they were able to correctly predict the proteins it was trained on? And maybe a few others?
It'd be kind of embarrassing if it couldn't.
AIs need huge data sets to train on after all, and I rather doubt we have have a library of known proteins large enough to provide both a viable training dataset AND enough of a testing dataset to have any real confidence in its accuracy.