Proteins are one of the most essential components that carry on various crucial functions in our bodies. These proteins are made up of amino acids, which arrange themselves in unique patterns and different shapes of folding sequences that have remained one of the biggest mysteries in the field of science. Scientists believe that many diseases that are incurable today can be treated if the exact structure of the core proteins involved in their pathophysiology can be determined. New drugs can be manufactured if the scientists understand the exact structure of a three-dimensional protein unit involved with the receptors, causing a particular disease.
For more than fifty years, scientists and researchers have been trying to find a close enough picture of a protein structure. So far, they have only discovered and identified 200 million proteins but their exact structures and how they fold into different sequences still remained a mystery. This research requires extremely dedicated scientists, a lot of money, and real hard work, which these people have put through all these years. Now, the Artificial Intelligence lab owned by Google, DeepMind seems to have made a leap in this direction and has provided the missing part of this grand puzzle.
DeepMind’s latest AI Program, AlphaFold has predicted how proteins fold into their unique three-dimensional shapes
As mentioned above, protein folding and its structure have been a huge missing link to understand the core of many diseases, and hence, their cures were not found as yet. However, DeepMind, the famous AI group which is popular for showing superhuman skills for various games has now made a breakthrough in science through its AlphaFold AI program.
DeepMind’s researchers trained their algorithm using a public database that contained around 170,000 protein sequences and their shapes. This algorithm training took several weeks with a lot of computing through almost 200 graphics processing units.
Once the algorithm was trained, DeepMind entered AlphaFold in the Critical Assessment of Protein Structure Prediction, or Casp, an event that occurs every two years in which different scientific teams from around 20 countries participate to predict the shape of a set of 100 proteins from the sequence of amino acids with the help of their computers. While these teams are predicting the shapes of proteins, their 3-D structures are created in the labs by biologists using basic techniques like X-ray crystallography and NMR spectroscopy. These efforts are made to locate the position of every atom in the protein molecular structure.
Scientists from Casp then compare the predictions from various teams with the 3-D structures that are worked out in the lab through different experimental methods. Once these results are compared, they are evaluated on the basis of a metric that Casp uses, called the Global Distance Test to assess the accuracy of those predictions on a score of 1-100.
Now, when DeepMind’s AlphaFold was also taking part this year, and to everyone’s astonishment, it achieved a score of 90 on the Casp Global Distance Test, which was comparable with the lab scores.
AlphaFold not only performed better than other computing programs, but it also attained an accuracy level that scientists achieved through extremely meticulous and time-consuming lab work.
AlphaFold’s score for the hardest proteins fell down a bit to 87 points, but the median score equivalent to all the experimental methods remained 90. All in all, AlphaFold was able to determine the structures of more than two-thirds of proteins accurately against lab experiments.
As per the experts, AlphaFold has done what researchers would have not been able to achieve at least for another ten years. AlphaFold’s deep learning methodology helps it assess the structure of a folded protein in the form of a spatial graph. The program then uses its knowledge to study further about the protein structure from the data available in a public worldwide database.
DeepMind has already begun giving out access to AlphaFold to various scientific researchers to help them determine the structure of extremely complex proteins.
Other scientists also want to look into this AI program to understand how it works and how accurate are the results?
DeepMind’s AlphaFold is garnering a lot of attention now from scientists all around the world who have invested a major portion of their lives in various lab works and experiments to solve the mystery related to the protein structure.
However, there is still a long way to go, because this is just the beginning. The whole puzzle is yet to be solved. This will help with single proteins, but their integration and interaction with other proteins in the human body and how DNA and RNA strands interact with each other remains a mystery to be solved. However, scientists are hopeful because AlphaFold’s program has helped the understand one problem, which will open the way towards an understanding of the structures of extremely complex proteins too with accuracy. And how these complex protein units work with other proteins to run the mechanisms in our bodies will also be revealed with time through advanced applications developed on the principled of AlphaFold.
I’m Ava Paul, an experienced news website author with a special focus on the entertainment section. Over the past five years, I have worked in various positions of media and communication at World Stock Market. My experience has given me extensive knowledge in writing, editing, researching and reporting on stories related to the entertainment industry.
