DeepMind says it will release the structure of every protein known to science - MIT Technology Review
( July 22, 2021; MIT Technology Review )
Deepmind co-founder, Demis Hassabis has been interested in protein folding for more than 20 years. Now, the platform he helped to create has predicted protein structures for about 350,000 proteins and plans to publish them along with predictions for about 100 million more in coming months. Essentially, this will cover all known proteins. Being able to grab a protein structure straight from a database instead of needing to calculate it with AlphaFold will make life easier for biologists.
Back in December 2020, DeepMind took the world of biology by surprise when it solved a 50-year grand challenge with AlphaFold, an AI tool that predicts the structure of proteins. Last week the London-based company published full details of that tool and released its source code.
Now the firm has announced that it has used its AI to predict the shapes of nearly every protein in the human body, as well as the shapes of hundreds of thousands of other proteins found in 20 of the most widely studied organisms, including yeast, fruit flies, and mice. The breakthrough could allow biologists from around the world to understand diseases better and develop new drugs.
Read the rest from : DeepMind says it will release the structure of every protein known to science
Check the #penny4thoughts tag to find other active conversations.
That's incredible. It's just like making a computer program open source, so it fosters innovation instead of locking it behind a paywall. Most biotech corporations keep this sort of things behind closed doors, instead of sharing them. Who knows how many cures are out there because corporations don’t share information.
I agree that it's an amazing achievement. I'm especially impressed because I know that WorldCommunityGrid has been working on protein folding since about 2004. For Google to say that they'll do 100 million proteins in a few months is stunning to me.
Releasing it to the public reminds me of Craig Ventner in the late '90s or early 2000s. If I recall correctly, part of his stated reason for sequencing the human genome was to make it public before other enterprises were able to patent it.
Ever since I started learning about epigenetics, I've been fascinated by the role of proteins in our fundamental cell function. This is a huge step with broad implications!
They have a lot of work to do then, considering how vast the number of distinct proteins are. As long as the AI's prediction of these protein are fairly accurate, then it's all good.
I read in the article that it's accurate to the level of an atom, which is detailed enough for drug development.
Yeah, but that is around 36%.
If the consistent accuracy remains true for the yet-to-be verified percentage, then we are in for a revolution.
It goes on to note that about 1/3 of the proteins in the human body are "floppy"; they only develop structure after attaching to other proteins. It sounds like it does leave about 1/3 with incorrect predictions, but evidently the predictions provide enough information for biologists to predict the function of a protein! What a resource!
And best of all, this company is launching all this research and its database for free, so that any researcher or laboratory can use it.
The world of biology perseveres in research until obtaining achievements, as in the specific case of proteins and their structure. Biologists are very consistent in their research. It is commendable. Thank you for such a good topic, I will delve further into it.
With all these amounts of proteins, they will generate a large database, which has to be very precise since it will be the basis for the development of drugs.