AI has taken over the world: from answering Biology questions, to writing emails to your boss, it has cemented itself in our daily lives. Although it has now been solved, Pablo Echenique – a famous physicist – called AI “the most important yet unsolved issue of modern science. The latest breakthrough in this science field came from AlphaFold AI, a technology that began to be developed in the 2010s. It now has become so advanced that it can predict protein structures with an astonishing 92% accuracy rate [1], which can now be utilized in saving millions of lives from detrimental illnesses. Firstly, we need to understand how this AI can predict these models and and how it’s different from other methods, and nextly,, how it can be used to save lives.
To start, the problem was having to predict the 3D structure of a protein using only its amino acid building blocks. This was very important advancement in the science community as trying to do so before was very inefficient and inexpensive. One way it was done before was X- Ray Crystallography which involved taking a protein crystal and shooting X-Rays at it in order to figure out its diffraction pattern. This then was given to a computer to analyze the data, a process that could take months, or even years. [2] It predicted close to 10,000 protein structures, compared to the 214 million that AlphaFold found. [3] AlphaFold, like any other AI, needed to be trained in order to predict these sequences of proteins, and the company behind it, DeepMind, used publicly available protein structures.. Once the AI was coherent and trained on these datasets, developers then gave only the amino acids leaving the AI to figure out what the 3D protein looked like. This combined with many errors and tweaks led to the first AI that was capable of predicting protein structures. This not only was a huge breakthrough in the scientific community, but also impacted the pharmaceutical industry and medical field.
The reason Pablo Echenique called this one of the most “important” issues is due to the sheer number of lives it can help. The past renderings of 3D protein structures were used to create medicines and treatments for them. For example, it helped figure out treatments for diabetes, sickle cell disease, breast cancer, etc. The structure of a protein is so important due to how treatments can be focused on “site-directed mutations”, changing or even hindering its function. AI also can help find a way to stop illnesses from binding to human cells in the first place. [4] This increase in knowledge helps researchers find ways to cure diseases thought to be nearly impossible to do so. A major example of the AI helping find treatments is Parkinson’s, which deteriorates the brain, limiting muscle movement, balance, and coordination. DeepMind estimates that 10 million people globally could be aided with this treatment alone. [5] This does not account for the many other diseases like Chagas disease that AlphaFold may help find treatment for [1].
Now, as AlphaFold continues to astonish many researchers with its huge potential and breakthroughs, it shows AI’s development and impact in our modern society. As times progress, AI will continuously help us find new ways to further our understanding of the world. AlphaFold is now revered highly for the knowledge it provides and millions of lives that it will not only impact but help save.
Bibliography
AlphaFold Protein Structure Database, https://alphafold.ebi.ac.uk/. Accessed 17 December 2023.
Buntz, Brian. “7 case studies highlighting the potential of DeepMind’s AlphaFold.” Drug Discovery and Development, 28 July 2022, https://www.drugdiscoverytrends.com/7-ways-deepmind-alphafold-used-life-sciences/. Accessed 17 December 2023.
Darnell, Steve. “Why Structure Prediction Matters | DNASTAR.” dnastar, https://www.dnastar.com/blog/protein-analysis-modeling/why-structure-prediction-matters/. Accessed 17 December 2023.
Komander, David. “Targeting early-onset Parkinson’s with AI.” Google DeepMind, 21 September 2022, https://deepmind.google/discover/blog/targeting-early-onset-parkinsons-with-ai/. Accessed 17 December 2023.
“X-ray Protein Crystallography.” Physics LibreTexts, 8 November 2022, https://phys.libretexts.org/Courses/University_of_California_Davis/UCD%3A_Biophysics_200A_-_Current_Techniques_in_Biophysics/X-ray_Protein_Crystallography. Accessed 17 December 2023. Accessed 17 December 2023.