The program cracks the scientific problem in stunning advance to comprehend the mechanism of life
After attaining fame on its superhuman achievement at playing games, the artificial intelligence group DeepMind has solved a momentous scientific challenge that has perplexed scientists for five decades.
By its newest AI program, AlphaFold, the organization and research laboratory proved that it can forecast how proteins fold into 3D shapes, an extremely complex process that is the key to realize the biological machinery of life.
Autonomous scientists think that the development would facilitate researchers tease apart the processes that drive some diseases and make way for designer medicines, more nutritious crops and green enzymes which can combat plastic pollution.
DeepMind opined that it had initiated work with a few scientific groups and would concentrate on malaria, sleeping sickness and leishmaniasis (a parasitic disease) at the beginning.
Demis Hassabis, DeepMind’s founder and chief executive feels that it marks an exciting moment for the field. Currently, these algorithms are becoming mature enough and robust enough to be applied to challenging scientific problems.
Venki Ramakrishnan, the president of the Royal Society, termed the work as a stunning advance that took place prior to the prediction of many people in the field decades before.
DeepMind is famous for organizing of human-trouncing programs that attained predominance in chess, Go, Starcraft II and old-school Atari classics. But superhuman gameplay wasn’t ever the primary objective. Rather, games offered a training ground for programs that, once powerful enough, would be set free on real-world problems.
Protein folding has been a grand challenge in biology for 50 years. An arcane form of molecular origami, its significance is hard to overstate. Most biological processes are circled around proteins and a protein’s shape regulates its function. Once researchers understand how a protein folds up, they can begin uncovering what it does. How insulin controls sugar levels in the blood and how antibodies to fight coronavirus are both dictated by protein structure.
Scientists have recognized more than 200m proteins but structures are known for only a few of them. Conventionally, the shapes are invented through meticulous lab work that may consume years. Though computer scientists have made progress on the challenge, concluding the structure from a protein’s makeup is no mean achievement. Proteins are chains of amino acids that can twist and bend into mind-boggling types of shapes: a googol cubed, or 1 followed by 300 zeroes.
To figure out how proteins fold, researchers at DeepMind trained their algorithm on a public database featuring around 170,000 protein sequences and their shapes. Running on the equivalent of 100 to 200 graphics processing units, by contemporary standards, a modest amount of computing power, the training took a few weeks.
DeepMind put AlphaFold through its paces by entering it for a biennial Protein Olympics called Casp, the Critical Assessment of Protein Structure Prediction. Freshers to the international competition are given the amino acid sequences for about 100 proteins and challenged to work them out. The results from teams that apply computers are compared with those based on lab work.
AlphaFold not only outperformed other computer programs but reached a precision comparable to the arduous and time-consuming lab-based methods. After it was analyzed across all ranked proteins, AlphaFold had a median score of 92.5 out of 100, with 90 being the equivalent to experimental methods. For the hardest proteins, the median score descended, but only marginally to 87.
Hassabis said DeepMind had begun working on giving researchers access to AlphaFold to assist in scientific research. Andrei Lupas, the director of the Max Planck Institute for Developmental Biology in Tübingen, Germany, said he had already employed the program to fathom a protein structure that scientists were stuck on for a decade.
Janet Thornton, a director emeritus of EMBL’s European Bioinformatics Institute near Cambridge, who was not a part of the work, said she was excited to know the results.
John Jumper, a researcher on the team at DeepMind, said that they didn’t know until they saw the Casp results how far they had pushed the field. It is not the end of the work, though. Future research will work on how proteins incorporate to form larger complexes and how they communicate with other molecules in living organisms.