-
Categories
-
Pharmaceutical Intermediates
-
Active Pharmaceutical Ingredients
-
Food Additives
- Industrial Coatings
- Agrochemicals
- Dyes and Pigments
- Surfactant
- Flavors and Fragrances
- Chemical Reagents
- Catalyst and Auxiliary
- Natural Products
- Inorganic Chemistry
-
Organic Chemistry
-
Biochemical Engineering
- Analytical Chemistry
- Cosmetic Ingredient
-
Pharmaceutical Intermediates
Promotion
ECHEMI Mall
Wholesale
Weekly Price
Exhibition
News
-
Trade Service
。 DeepMind's deep learning techniques for determining the shape of proteins in 3D could revolutionn biology. The blue color in the figure is the protein structure predicted by the computer, and the green color is the experimental verification result, and the similarity between the two is very high. (Photo: DeepMind)if you liken organisms to a construction site, then protein is a brick on a site. There are thousands of different proteins in the body, each containing dozens of amino acids, the order of which determines the shape and function of the protein. "Structure as Function" is the therm of molecular biology, and if its structure can be introduced according to the amino acid sequence of proteins, it will help people to speed up the understanding of cell composition and operation, and the development of some new drugs can be advanced more quickly.
For too long, experiments have been needed to determine complete protein structures, such as X-ray crystallism and cryoscopes, which often take months or even years, and less than 200,000 of the 200 million proteins that have been discovered in humans have been parsed.
now, artificial intelligence has the ability to give accurate predictions, even for days or even half an hour. DeepMind's AlphaFold program recently stood out among more than 100 teams at the Protein Prediction Structure Challenge CASP. One of CASP's rules of the game is that the protein structure predicted by the contestants must have been experimentally validated but not published publicly. The predicted results are tested anonymously by experimental methods, and the higher the similarity between the two, the higher the score.
competition, DeepMind's AlphaFold combines deep learning with a stress control algorithm and applies it to structural and genetic data, and the deep learning network trains with 170,000 proteins that are currently known to have been parsed. Combined with the physical structure and geometric constraints of protein folding, AlphaFold can predict the sequence structure of the target protein - even including proteins wedged into the cell membrane, which is key to understanding many human diseases.
but AlphaFold wasn't perfect, and alphaFold didn't get a high score when predicting a protein structure of 52 small repeating fragments. DeepMind has now published details of the first version of AlphaFold for other researchers to replicate. DeepMind's research and development team says it will continue to train AlphaFold to better analyze more complex protein structures.