DeepMind’s and protein coagulation problem

DeepMind’s AI gets ahead scientists: Neural networks are approaching the biological clue to the “protein clotting problem”

Every 2 years in one of the selected countries, international competitions Critical Assessmant of Structure Prediction are held, where hundreds of chemists and scientists propose a solution to the “protein clotting problem”. What task do scientists set for themselves at this event?

The main goal is to accurately predict the physical structure of the protein in three dimensions. If it is possible to determine its shape, then pharmacists will be able to develop new medicines based on knowledge of the attachment of other molecules to the protein. Finding the answer will simplify the development of new drugs for the treatment of a number of diseases.

The paradox is that even the winners of past competitions do not know the valid answer to the solution of this question.

At the last competition, to the entire scientific world’s amazement, the victory was get by the company DeepMind (Google’s AI laboratory). Neural networks outstrip academicians and made a step forward to solve the problem of protein clotting.

How did they manage to do it?

DeepMind has focused on deep AI learning to change the science of medical supplies discovery. Making forecasts using neural networks will activate a variety of aspects in solving pharmacology problems that people have previously taken on.

The latest lab projects are aimed at this.

In early 2016, the DeepMind company using the AI system was able to beat people in the game of Go. After the success, the British researchers continued to search for new tasks for neural networks and chose the problem of “protein clotting”.

With the help of the game, it was possible to imitate a scientific task and develop an AI system that learnt on its on. The neural network simultaneously analyzed thousands of proteins and predicted the shape of others. The technology is similar to the face recognition system in photos that Facebook uses. Academics at Critical Assessmant of Structure Prediction used similar techniques, but DeepMind became the winner with a wide margin. The AI was able to predict the results 2 times more accurately in comparison with chemists.

What does the victory of DeepMind’s AI?

Future research breakthroughs for biochemistry will depend directly on the machines and their computing power. If before universities and pharmaceutical companies could not use such significant resources, then in the future Google will be able to provide cloud computing services. The fall in prices for the computing power of AI is already beginning to attract the global scientific community.

The task of studying the behavior and form of proteins will allow scientists to discover new proteins, which will give measurable results for doctors and patients. But it still takes time.

Rating
( No ratings yet )
intelligentmd-news.net
Leave a Reply