I think that almost all of you are familiar with a new shocking AI technology called Dall-e, which was introduced quite recently. This is a platform that generates images by just specifying what you wish to view. Social media sites are now crowded with these surprisingly detailed, often photorealistic images created by this or similar technologies. However, some scientists perceive it as more than just a means of creating pictures. They see it as a way to treat various diseases, such as cancer or flu.
Recently, by using these modern AI technologies, scientists have started generating blueprints for new proteins – tiny biological mechanisms that play a significant role in our bodies’ operation, ranging from digesting food to moving oxygen through the bloodstream. Although these proteins are produced naturally in our bodies, researchers are still striving to improve the ability to fight diseases and do things that our bodies can not produce on their own.
For more than 30 years, David Baker, the head of the Institute for Protein Design at the University of Washington, has worked to develop artisanal proteins. He and his colleagues have established this was feasible by 2017. However, they did not anticipate how the emergence of new AI technologies would radically speed up this task, cutting the period of time required to produce new blueprints from years to only a few weeks.
Proteins are made up of long chains of chemical components that then twist and fold into three-dimensional structures. Recent research from AI labs like DeepMind, which is owned by Alphabet, has demonstrated that neural networks can successfully predict the three-dimensional shape of any protein in the body based only on the smaller compounds it contains.
Nowadays researchers are taking a step further by creating blueprints for totally new proteins that do not exist in nature, by using AI systems. The objective is to develop proteins that adopt highly specific shapes. A particular shape can perform a certain function, such as preventing the COVID-19 virus. Researchers can provide a rough description of the protein they want, then a diffusion model can generate its three-dimensional shape. However, scientists still need to test it in a wet lab with actual chemical compounds to make sure it functions as expected.
On the one hand, some experts take this innovation with a grain of salt. Frances Arnold, a Nobel laureate, comments it as “Just a game”. He states that what really matters is what a generated structure can actually do.
On the contrary, Andrei Lupas, an evolutionary german biologist, is convinced that it will change medicine, research and bioengineering. “It will change everything”. AlphaFold has helped him to find the structure of a protein he was tinkering with for almost a decade.
Personally I agree with a majority of researches and assume that AI is a tool for exploring new innovations that scientists could not previously think on their own.
References:
