AI in Biology: Predicting Protein Structures

Proteins are essential structures for life. They make up everything, from our bones to our skin and muscles. These proteins are assembled from hundreds or thousands of smaller amino acids linked together by peptide bonds. These amino acids differ based on the composition of their side chains, which are able to interact with the side chains of other amino acids in a multitude of different ways. The complex interactions between the many amino acids within the chain are what fold and twist the chain into the final shape and conformation of the protein, which determines its function and efficacy. Because of the innumerable ways a given protein chain can fold, biologists could only for the longest time determine the molecular structure of proteins using experimental data and months or years of effort.

Introducing: AlphaFold

Finally, in late 2020, DeemMind released AlphaFold, which would become the game-changing tool that innovates the use of AI in biochemistry. Its most recent model, AlphaFold 3, was released on May 8 2024. This model determines the 3D structure of proteins using both evolutionary approaches, that focus on evolutionary history, pairwise correlation and homology; and physical approaches, based on thermodynamic or kinetic simulations of the molecular interactions within proteins. This method has proven to be highly accurate, having a median backbone accuracy of 0.96 Å r.m.s.d.95 (meaning that the median distance between the predicted and actual location of Carbon atoms for 95% of the protein structure is 0.96 Å, or 0.096 nanometres). This is a great improvement from the next best performing method that has a median backbone accuracy of 2.8 Å r.m.s.d.95.

Why should we know the molecular structure of proteins?

It’s important to understand the molecular structure of proteins because they are directly related to its function. The order of amino acids within an amino acid chain determines the final conformation and hence how it interacts with other molecules to function. If we are able to determine his confirmation given only the amino acid sequence, we would be able to better understand how proteins function. That opens whole new worlds as we would be able to use this information to track and target certain proteins associated with diseases and synthesize new materials and catalysts for industrial and daily use.

How is AlphaFold being applied?

Here are some specific cases that Google DeepMind, the developer of AlphaFold, shares on their website:

1. AlphaFold is being used to block the resistance mechanism of certain bacteria that are immune to antibodies. By targeting enzymes within the bacteria, we would be able to more easily treat the millions of antibiotic-resistant infections using existing antibiotics.

2. AlphaFold is being used to better understand the genetic causes of osteoporosis. In that way, we may recognize people with a higher risk of developing the disease, allowing for earlier diagnosis. This would be an improvement from current tools that recognize osteoporosis only after complications have begun.

3. AlphaFold is being used to study extinct life forms, by applying its knowledge of proteins found in organisms today to proteins found in the egg of an ancient bird. This allows us to more accurately analyze the appearances and tendencies of extinct creatures, and gives us a hint as to why they went extinct.

AlphaFold is just one of many developments being made in the field of health and science. If you’re interested in learning more about how technology is improving in our healthcare and research fields, feel free to read more articles from our ‘Technology’ section!

References

About protein. Why is the study of proteins important? | Protein Crystal Growth on the International Space Station | JAXA Human Spaceflight Technology Directorate. (n.d.). https://humans-in-space.jaxa.jp/protein/en/public/about/why.html

Callaway, E. (2024, October 9). Chemistry Nobel goes to developers of AlphaFold AI that predicts protein structures. Nature News. https://www.nature.com/articles/d41586-024-03214-7

Formosa, M. (2022, September 27). Fighting osteoporosis before it starts. Google DeepMind. https://deepmind.google/discover/blog/fighting-osteoporosis-before-it-starts/?_gl=1%2Amj0lzn%2A_up%2AMQ..%2A_ga%2AMTUwOTY2NTA2NS4xNzQzMzA2NTg0%2A_ga_LS8HVHCNQ0%2AMTc0MzMwNjU4NC4xLjAuMTc0MzMwNjU5Ny4wLjAuMA..

Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S. A., Ballard, A. J., Cowie, A., Romera-Paredes, B., Nikolov, S., Jain, R., Adler, J., … Hassabis, D. (2021). Highly accurate protein structure prediction with alphafold. Nature, 596(7873), 583–589. https://doi.org/10.1038/s41586-021-03819-2.

Service, R. F. (2020, November 30). ‘The game has changed.’ ai triumphs at solving protein structures | science | AAAS. Science. https://www.science.org/content/article/game-has-changed-ai-triumphs-solving-protein-structures

Solving the mystery of how an ancient bird went extinct. Google DeepMind. (2022, September 22). https://deepmind.google/discover/blog/solving-the-mystery-of-how-an-ancient-bird-

went-extinct/?_gl=1%2A1107d9%2A_up%2AMQ..%2A_ga%2AMTI4NDc3MDEwMS4xNzQzNTY5NTk3%2A_ga_LS8HVHCNQ0%2AMTc0MzU2OTU5Ni4xLjAuMTc0MzU2OTYwOC4wLjAuMA..

Mitchell, M. (2022, July 28). Accelerating the race against antibiotic resistance. Google DeepMind. https://deepmind.google/discover/blog/accelerating-the-race-against-antibiotic-resistance/?_gl=1%2Azdh7l1%2A_up%2AMQ..%2A_ga%2AMTUwOTY2NTA2NS4xNzQzMzA2NTg0%2A_ga_LS8HVHCNQ0%2AMTc0MzMwNjU4NC4xLjAuMTc0MzMwNjYwMi4wLjAuMA..

Veritasium. (2025, February 10). The Most Useful Thing AI Has Ever Done. YouTube. https://www.youtube.com/watch?v=P_fHJIYENdI&ab_channel=Veritasium.