Tremendous developments in the field of computer science open up multiple possibilities and due to it biologists can dream up 100% natural molecules in the nick of time instead of long weeks or even months. Recently, South Korea approved a COVID19 vaccine, constituted from new proteins developed by people. The medication is grounded on a spherical protein ‘nano-particle’ produced through a labor-intensive trial-and-error process.
AI defines the shape of almost all known proteins
AI instruments such as DeepMind’s software ΑlphaFold which can predict protein structures are actively applied for research purposes. In the past, this had been a challenging task with high likelihood of failures. Biologists and a plethora of organizations that are using AI for enzyme designing aim to create substances that can have a beneficial impact on human organisms, from getting rid of dangerous pathogens to curing illnesses. Among the organizations that are pursuing this purpose are Deep-Mind and Meta.
Designing proteins from scratch
A try to dream up the first enzymes dates back to the 1990s, when software under the name of Rοsetta was applied to research by Baker’s laboratory. The first proteins designed by AI in the majority of cases did not set up into the expected shape when produced in the laboratory. That’s why a new step was necessary to set the enzyme sequencing in such a way that it set up only into an expected form.
With ΑlphaFold and other AI programs, that slow process has become very fast. These tools were instructed to forecast the structural aspect of separate set of amino acids. But biologists saw that such software had also an ability to create sets of a range of enzymes that form functional interactions. At this point, they assured they were able to dream up enzymes that would assemble themselves into nano-particles of various forms and parameters. Such proteins would consist of a bunch of duplicates of one enzyme and resemble those on which the COVID19 medicine is grounded.
Baker’s laboratory isn’t the only organization using AI to create protein artificially. According to biοRxiv, there are more than forty AI programs that have been available recently that apply different methods.
The majority of these programs deal with the inverse fοlding algorithm issue: they determine sequencing type that is right to a certain composition, often applying methods used in picture-recognition software. Others are grounded on architectures close to that of natural language processing systems, which produce texts; however, in lieu of texts, the programs define protein sequencing.
Scientists at Baker’s lab are sure that designing proteins in the laboratory conditions is the conclusive analysis of their approaches. But not all biologists working on AI software for enzyme designing have hassle-free access to exploratory setups, and this is challenging for testing their findings, as finding a lab to make probing can be time-consuming.
Research will also be of high significance concerning dreaming up proteins for certain tasks. For this purpose, Baker’s lab shared several AI arrangements that make it possible for scientists to imbed a certain sequencing order in a new enzyme. These methods were useful in designing enzymes that catalyze certain responses:
- proteins able of joining other particles; and
- proteins that potentially can be applied in medicines against respiratοry viruses.