'Scientific Breakthrough': AI program solves decades-long ‘protein folding problem’

An artificial intelligence (AI) created in the UK has cracked decades-long 'protein folding problem’ after it mapped out their 2D structures and distinguished between the different types of proteins. For more than 50-years, scientists were unable to decode biology’s one of the biggest challenges that could help in the development of treatments and drug effectiveness. On November 30, however, a London-based Google-owned artificial lab DeepMind announced that its program AlphaFold was now able to detect many shapes of the proteins that could accelerate research in fields of drug design, treating diseases, and environmental sustainability. 

“Figuring out what shapes proteins fold into is known as the protein folding problem, and has stood as a grand challenge in biology for the past 50 years,” the AI lab informed in a release.

It added, that the large complex molecules composed of amino acids were comprised in all life forms’ supporting systems, including in COVID-19 disease-causing viruses SRAS-coV-2 and in cancer cells.

In a breakthrough, DeepMind’s biennial Critical Assessment of protein Structure Prediction (CASP) was able to detect how the proteins fold up by the thorough assessment of several experimentally determined protein structures. In total, there are approximately 200 million known proteins that exist. Researchers from the 14th Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP14) solved the 1994 riddle by creating an attention-based neural network system that was trained end-to-end to interpret the structures using the latest version of AI AlphaFold. 

Read: 'Super Planet' Discovered By Scientists Using Radio Telescope For The First Time

Read: Scientists Create Diamonds In Lab At Room Temperature By Mimicking Asteroid Collision

[Two examples of targets in the free modelling category. Credit: Deepmind site]

'Trained' on 170,000 protein structures

“AI system developed strong predictions of the underlying physical structure of the protein and is able to determine highly-accurate structures in a matter of days,” scientists informed in a release. They added, that AlphaFold could also predict “which parts of each predicted protein structure were reliable.” AlphaFold was trained on 170,000 protein structures, generated an average accuracy score of 92.4 out of 100 in prediction. “This computational work represents a stunning advance on the protein-folding problem, a 50-year-old grand challenge in biology,” said President of the Royal Society Venki Ramakrishnan. “It will be exciting to see the many ways in which it will fundamentally change biological research,” he added.

Read: Israeli Scientists Claim To Biologically Reverse Human Ageing Process For First Time Ever

Read: British Scientists 'fine Tuning' Process To Extract Oxygen From Moon Dust 'Lunar Regolith'