Surprisingly, it now turns out that artificial intelligences that 'paint', like Dali, can push the world of science forward
We regularly hear news about new artificial intelligence engines for creating texts and drawings, but it is easy to forget that artificial intelligence has an even greater potential: to create scientific knowledge and develop new technologies. Surprisingly, it now turns out that precisely 'painting' artificial intelligences, like Dali, can push the world of science forward.
Two biotech labs have recently begun using 'drawing' artificial intelligence to produce blueprints and diagrams of new proteins that never existed in nature. In one of the laboratories, the researchers continued and produced the new protein - and showed that it was able to perform the task they had set for it, better than any other protein designed so far. And of course each of the two laboratories launched software that allows anyone to sketch the proteins of the future.
But what does all this mean?
Let's start, of course, at the beginning.
Proteins are one of the most important building blocks of the body. They are both used as a building material and act literally as machines inside the body: they break down the food, help in the production of energy in the cell, allow the cells to maintain a rigid structure and generally play a role in every possible action in the cell. The proteins also allow the cells to react and 'talk' to each other, because the receptors on the surface of the cell are made of - you guessed it - proteins. And of course, viruses also use proteins to attach to cells and invade them, and bacteria release their own malicious proteins, which damage cells.
No wonder drug developers focus first and foremost on proteins when they want to find a cure for any disease. In some cases, they identify the protein that causes the disease, and then try to invent their own protein that will catch the little pest and paralyze it. In other cases, they try to find a protein that can kill bacteria or viruses, or perhaps connect to cancer cells and send a signal to the immune system that will come to eliminate them. Be that as it may, proteins are seen as a possible answer to countless diseases.
There is only one problem: it is very difficult to design new proteins.
Human proteins are themselves composed of a collection of Lego blocks: twenty different types of amino acids that can connect to each other to form a long strand, which sometimes contains tens of thousands of such building blocks. After the strand is formed, it immediately begins to curl around itself: the different acids attract others that are far away from them, repel others and form and break bonds. At the end of this folding process, the finished three-dimensional machine is obtained, which is the protein. All this happens in cells every minute, but to understand in advance how to create these machines, we still needed supercomputers - and even they were not enough to solve the folding problem of the more complex proteins.
Then came Dali.
Dall - Dall-E - is an artificial intelligence engine used to generate images and drawings. It works in a very strange way: it takes random white 'noise', gradually cleans and filters it, until an image that matches the user's desire is left behind. It is no wonder that one of the similar engines operating according to the same method is called Stable Diffusion. The chaos dissipates, and a stable image is obtained.
Two biotech labs have launched artificial intelligences that work on a similar principle to 'paint' proteins. Both start from a state of extreme disorder, by breaking down the bonds between the amino acids that make up the protein. After disassembly, the programs try to reassemble them into a protein in a certain final configuration that the user defines.
Humans need months and even years to design a new protein. The new artificial intelligences, on the other hand, are able to do this almost instantly. To demonstrate their motor capability, one of the laboratories - Generate Biomedicines - chose to produce proteins in the shape of all 26 letters of the Latin alphabet. Why? Why not. When it is so easy to produce a new protein, there is no reason to hold back.
Still, it must be admitted that proteins in the form of letters of the alphabet are not going to change the world of health tomorrow morning. But it is clear to everyone that this is only a demonstration of abilities. An artificial intelligence that today can produce a protein in the shape of the letter A, tomorrow morning will also be able to produce a protein that will attach to viruses and paralyze them, or a protein that will identify cancer cells and target them, or a protein that will attach to the fat in the arteries and help break it down. And the artificial intelligence will be able to draw all these for us.
But who guarantees us that the beautiful proteins she will sketch, we can really create in the laboratory?
At this point another artificial intelligence came into play, which the researchers ran on the beautiful blueprints of the first intelligence. The second intelligence was supposed to determine whether it is really possible to create the innovative proteins in the laboratory - and it determined that out of all the ideas produced, 55 percent of them can certainly be produced in reality.
The second lab - that of David Baker from the University of Washington - went one step further. The researchers practically created a new protein according to the inspiration and instructions provided by the artificial intelligence. It is not just a protein, but a creation with the potential to deal with an existing disease. They chose to produce a protein that is able to attach to the parathyroid hormone, which controls calcium levels in the blood. When the researchers tested the new protein, they discovered that it is able to bind to the hormone better and more strongly than any other protein discovered so far - including those given in drug treatments.
To illustrate the significance of this development, here is what Baker himself said in an interview with MIT Technology Review -
"We basically gave the model the hormone and nothing else, and told it to create a protein that would bind to it." He said, adding that, "He brought this plan to the protein out of absolutely nothing."[1]
And here is the way in which medicines will be developed in the future: by artificial intelligence that will be able to come up with ideas for new medicines in a few minutes, test them in computer simulations - and finally also choose the successful ones among them and move forward with them to clinical trials. The right to do such computerized experiments will not be reserved for big researchers in dental laboratories and pharmaceutical companies. Thanks to the computer models, it will also reach children in schools. Just as some of the children of the Internet will create algorithms, games and companies before they finish school, so there will soon be children who will develop the medicine of the future and propel it forward.
It is difficult for me to stop here without adding a word or two of warning. As easily as it would be possible to produce a good and beneficial protein, the children will also be able to produce nerve poison and drugs. And when the technology advances even further, they will also be able to produce real nanotechnological machines, which will travel through our bodies and do... things. Obviously, there is also the potential for harm in this technology, which allows us to play with the Legos of biology and reassemble them in any way we want. But the potential for good is so great - and I'm not ashamed to write here that it's endless - that it's hard for me to wait for the technology to reach the laboratories and even the masses. It will provide another powerful tool in the hands of scientists, and advance us another big step towards a future without diseases.
And a final quote from David Baker, who has been designing proteins in his lab for many years -
"Scientific progress happens in Kitoin - ups and downs. But now we are at the center of what can only be called a technological revolution."[2]
[1] https://www.technologyreview.com/2022/12/01/1064023/biotech-labs-are-using-ai-inspired-by-dall-e-to-invent-new-drugs/
[2] https://www.technologyreview.com/2022/12/01/1064023/biotech-labs-are-using-ai-inspired-by-dall-e-to-invent-new-drugs/
More of the topic in Hayadan:
