DeepMind’s AlphaFold AI Could Accelerate Drug Discovery
DeepMind’s AlphaFold is poised to revolutionize protein structure prediction, and its many real-world applications, through machine learning. Predicting the folded structure of proteins from their DNA has always been a difficult and time-consuming process. However, DeepMind’s recent breakthrough with AlphaFold could significantly reduce the amount of time required to produce an accurate prediction.
The Context:
In 2020, AlphaFold participated in Critical Assessment of Structure Prediction (CASP) 14, a challenge during which teams attempt to predict the structure of folded proteins. During this challenge, AlphaFold outperformed 100 other teams with several predictions that closely resembled those produced by the benchmark experimental methods, including X-ray crystallography and cryo-electron microscopy (cryo-EM). Even more groundbreaking, it was the first time that a team had used machine learning and AI to produce an accurate prediction within days.
The Science:
To achieve the breakthrough, DeepMind built on the work done by scientists during the last 50 years and incorporated ideas from the fields of machine learning, physics, and biology. First, they decided to treat folded proteins as “spatial graphs” and use an algorithm that strives to determine the structure of the graphs. At the same time, the algorithm utilizes “evolutionarily related sequences, multiple sequence alignment, and a representation of amino acid residue pair” to review the graph it is in the process of building. By repeating this process, AlphaFold learns to predict accurate protein structures in just days.
Impact:
This development is potentially very useful for the medical community. By greatly expanding the number of known protein structures, possibly into the hundreds of millions from the hundreds of thousands, AlphaFold could open up many avenues for research and treatments for debilitating illnesses such as Alzheimer’s and Parkinson’s. DeepMind has already illustrated how this technology could assist epidemiology. During the early stages of the COVID-19 pandemic, AlphaFold made several predictions about the shape of several SARS-COV-2 proteins, one of which was eventually confirmed by traditional methods.
Learn More:
If you would like to learn more about the ways that AI and data science are transforming the medical field, check out the Ai+ Training Platform. With a monthly subscription, you can watch on-demand training videos about the latest advances in data science topics and tools, including the one below.
- Natural Language Processing Case-studies for Healthcare Models│Veysel Kocaman, PhD│Senior Data Scientist│John Snow Labs
- Health AI: What’s Possible Now and What’s Hard│Suchi Saria, PhD│Director Of The Machine Learning And Healthcare Lab│John Hopkins University
For training sessions on essential data science topics and tools check out Dr. Jon Krohn’s 14-part machine learning mini-bootcamp, The Foundations for Machine Learning, free with an annual Premium subscription.
Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform.
