All About AlphaFold, the Protein Structure Predictor

Proteins play a major role in a multitude of life processes & predicting the structure, and consequently the function, of a protein is a nuanced endeavor. Keep reading to learn more about AlphaFold, a new technology that accurately predicts protein structures, “solving” the protein folding problem.

Sections:
  1. Why is Predicting Protein Structures Important?
  2. DeepMind’s AlphaFold AI Tackles the Protein Folding Problem
  3. How did DeepMind Approach the Protein Folding Problem?
  4. Applications
(TL;DR, Glossary, and a Video Resource at the bottom 😊)
Approximate Read Time: 4 minutes!

Why is Predicting Protein Structures Important?

Proteins are incredibly important. From understanding viruses to developing drug treatments to even studying proteins that break down waste, it is undeniable that proteins are an integral part of our world & it’s important to understand how they do what they do!

Predicting the ways in which proteins fold and studying their final structures are crucial, as protein structure tells us about how the proteins function (simply worded, structure = function).

3D model of a protein

In the scientific community, the “protein folding problem” is known as the challenge of accurate protein structure prediction. 

Why is protein structure prediction so challenging?

There are an extremely large amount of ways a protein could theoretically fold. Additionally, there are more than 200 million proteins in the global database & many of these protein structures are unknown. Without an accurate protein-prediction technology, protein structures must be determined experimentally, which is a long and tedious process that requires expensive equipment.

Bottom-line? An accurate protein structure prediction technology has been long sought after, as it would make the process of understanding how proteins work much more efficient.


DeepMind’s AlphaFold AI Tackles the Protein Folding Problem

DeepMind is a multidisciplinary team of scholars who work to build safe AI systems that advance scientific analysis. DeepMind created the AlphaFold AI to tackle predicting protein structures.

AlphaFold achieves protein structure prediction with significantly higher levels of accuracy than previous protein structure prediction software! AlphaFold is also able to determine the structures of proteins within a few days — a feat unachievable beforehand.

AlphaFold was tested through the Critical Assessment of protein Structure Prediction (CASP), which is a community-run analysis of predictive techniques. DeepMind, the creators of AlphaFold, used their software to predict the structure of a novel protein using the protein’s amino-acid sequence. Then, experimental data were collected to determine the structure of the protein, which was then compared to AlphaFold’s predicted model.

To assess the accuracy of AlphaFold’s prediction model, CASP uses the Global Distance Test (GDT), which operates on a scale from 0-100 with 100 signifying a perfect prediction. The AlphaFold prediction obtained a median score of 92.4 and a mean score of ~88. 

According to CASP, a score of at least 90 on the GDT signifies a model compatible with the experimental results!

“We have been stuck on this one problem — how do proteins fold up — for nearly 50 years. To see DeepMind produce a solution for this, having worked personally on this problem for so long and after so many stops and starts wondering if we’d ever get there, is a very special moment.”

Professor John Moult, Co-Founder and Chair of CASP, University of Maryland

How did DeepMind Approach the Protein Folding Problem?

The researchers at DeepMind thought of a folded protein as a “spatial graph,” with nodes and edges used to represent the physical interactions within the proteins. The scientists then created an attention-based neural network system to build the structure of the “spatial graph” protein using evolutionarily related sequences and the given protein’s amino acid sequence.

An overview of the main neural network model architecture of AlphaFold

The AlphaFold AI system was trained using publicly available data of ~170,000 known protein structures. 

“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.”

Sundar Pichai, CEO, Google and Alphabet

Applications

AlphaFold can be used to better understand diseases (such as COVID-19) and better develop drug treatments. 

“AlphaFold is a once in a generation advance, predicting protein structures with incredible speed and precision. This leap forward demonstrates how computational methods are poised to transform research in biology and hold much promise for accelerating the drug discovery process.”

Arthur D. Levinson, PhD, Founder & CEO Calico, Former Chairman & CEO, Genentech

Furthermore, AlphaFold has really demonstrated the ways in which AI can help progress scientific research!

What are your thoughts on AlphaFold?

Thank you for reading this piece. I hope you enjoyed it & will further ponder the knowledge it brings to light. I encourage you to leave a comment and/or start a discussion in the section below! 

Feel free to contact me at brainsproutblog@gmail.com and follow the blog @brainsproutblog on Instagram or Twitter for updates!


TL;DR
  • The “protein folding problem” is known as the challenge of accurate protein structure prediction. An accurate protein structure prediction technology has been long sought after, as it would make the process of understanding how proteins work much more efficient
  • AlphaFold achieves protein structure prediction with significantly higher levels of accuracy than previous protein structure prediction software! AlphaFold is also able to determine the structures of proteins within a few days — a feat unachievable beforehand
  • AlphaFold can be used to better understand diseases (such as COVID-19) and better develop drug treatments
Glossary
  • AI: artificial intelligence; intelligence demonstrated by machines
  • Amino acid: any of a group of organic molecules that consist of a basic amino group (―NH2), an acidic carboxyl group (―COOH), and an organic R group (or side chain); the building blocks of proteins
  • Protein: any of various naturally occurring extremely complex substances that consist of amino-acid residues joined by peptide bonds, contain the elements carbon, hydrogen, nitrogen, oxygen, usually sulfur, and occasionally other elements (such as phosphorus or iron), and include many essential biological compounds (such as enzymes, hormones, or antibodies)
Video Resource
Read More (sources)!
  1. https://scitechdaily.com/major-scientific-advance-deepmind-ai-alphafold-solves-50-year-old-grand-challenge-of-protein-structure-prediction/ 
  2. https://scitechdaily.com/artificial-intelligence-solution-to-a-50-year-old-science-challenge-could-revolutionize-medical-research/ 
  3. https://www.nature.com/articles/d41586-020-03348-4 
  4. https://deepmind.com/blog/article/alphafold-a-solution-to-a-50-year-old-grand-challenge-in-biology 
Image Sources
  1. https://www.pinclipart.com/maxpin/ihJiobm/
  2. https://www.nature.com/articles/d41586-020-03348-4
  3. https://scitechdaily.com/major-scientific-advance-deepmind-ai-alphafold-solves-50-year-old-grand-challenge-of-protein-structure-prediction/ 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Create your website with WordPress.com
Get started
%d bloggers like this: