This AI Thinks Like a Real Chemist After Solving a Protein-Folding Problem

Artificial intelligence has gone a long way in the field of science. Scientists used to incorporate it in their studies, including the analysis of molecular components in a living body.

Through deep learning, AI has become more progressive in helping the experts with testing organic compounds and even designing a building block.

Recently, a chemistry expert has found out that an AI can learn how to explore protein folding. Later, it managed to solve the issue behind this process.

AI Learns How to Solve Protein-Folding

Protein is an integral part of all biological creatures since it helps the cells carry out many processes, including food digestion and molecule transportation.

However, in order to perform its tasks, a protein should first undergo a process dubbed "protein folding," which involves twisting and bending the protein chain into a 3D structure.

According to The Next Web (TNW), it's important to know the proper arrangement of amino acids so that misfolding won't happen. In case it takes place, the affected proteins can lead to disease.

The protein-folding problem has become a huge puzzle in the minds of scientists in the past 50 years. Not even one of them got close to solving it.

However, when Alphabet's AI subsidiary DeepMind started its Alphafold program in 2016, the experts saw the light when they established a protein data bank.

AlphaFold2 Learns Chemistry

This training set is said to house various structures from hundreds of proteins. AlphaFold has been successful so far as it succeeded in solving the protein-folding puzzle.

This scientific breakthrough is regarded to be game-changing in the field of AI. Speaking of which, AlphaFold2 has been an important tool in modeling the protein combination.

Although it's not designed to perform protein interactions predictions, it has created a huge name in chemistry.

To be specific, this AI discovered an interesting thing about fluorescent proteins. It found out how the amino acids glow.

As per TNW, AlphaFold2's ability to learn chemistry can be attributed to the protein data bank training set, as well as the sequence alignments.

In general, this discovery will be a huge help for the chemists who will use deep learning in other aspects such as facial recognition, consumer spending, and more.

Physicists Teach Physics to AI

In another story from Sci-Tech Daily, Duke University researchers used machine-learning algorithms. The experts have managed to predict what metamaterial holds. What's more, is that they did it with high accuracy.

Additionally, the team has discovered that upon teaching physics to artificial intelligence, the AI immediately created new discoveries on its own.

This article is owned by Tech Times

Written by Joseph Henry