AlphaFold — Is it a Revolutionary Scientific Breakthrough?

Ambuj Agrawal
DataSeries
Published in
3 min readDec 13, 2020

Excluding water and fat, the human body is made up almost entirely of protein. Protein is the main component of muscles, bones, organs, skin, and nails. Proteins are commonly referred to as building blocks of the human body. The function that a protein performs is largely related to the shape it is folded in, and being able to predict the shape of a protein has the potential to solve some of the world’s biggest challenges like developing effective treatments for diseases and finding suitable enzymes for certain tasks.

Figuring out what shape a protein is folded in is known as the ‘protein folding problem’, a 50-year-old challenge that biologists have been striving to solve. AI can help researchers in creating a solution to this long-standing problem, showcasing the impact that AI can have on scientific discovery and some of the most fundamental problems that have the potential to revolutionize the world.

The DeepMind team trained deep neural networks to identify properties of the protein, given its genetic sequence. Two properties were predicted — the distances between pairs of amino acids, and the angles between bonds that connected the amino acids.

One deep neural network was used to predict a distribution of distances between every pair of residues in a protein. A score that estimates how accurate a proposed protein structure was then obtained from these probabilities. A different neural network estimated how close the proposed structure is to the correct structure, using all the distances.

Possible structures that matched the predictions of the model were then searched for, using the scoring functions. The first method replaced pieces of a protein structure with new protein fragments repeatedly, building on a common method in biology. A generative neural network was trained to invent new fragments, using which the score of the predicted structure of the protein was improved continuously. The second method used gradient descent (backpropagation in case of neural networks), which is a common optimization algorithm in machine learning. AlphaFold’s analysis matched almost perfectly with the CASP analysis for two-thirds of the proteins, compared to about 10% from the other teams.

“Determining a single protein structure often required years of experimental effort,” said Janet Thornton, director emeritus of the European Bioinformatics Institute and one of the pioneers of using computational approaches to understanding protein structure. “A better understanding of protein structures and the ability to predict them using a computer means a better understanding of life, evolution and, of course, human health and disease.”

“This is an incredible AI-powered breakthrough in protein folding, which will help us better understand one of life’s most fundamental building blocks. This huge leap forward from DeepMind has immediate practical implications, enabling researchers to tackle new and difficult problems, from future pandemic response to environmental sustainability”, said Sundar Pichai, CEO of Alphabet and Google, the parent company of DeepMind.

AlphaFold is indeed a Scientific breakthrough and he pace at which Artificial Intelligence is evolving, with it finding applications and breakthroughs in some of the most fundamental and revolutionary aspects of the modern world, AI is sure to be a frontrunner in driving humanity forward in the years to come.

--

--

Ambuj Agrawal
DataSeries

Ambuj is a published author and industry expert in Artificial Intelligence and Enterprise Automation (https://www.linkedin.com/in/ambujagrawal/)