Go deeper - join us!
Before we dive into this, I want to give a very small (and admittedly simple) overview of proteins and their significance in biological science. In short, proteins are essential building blocks of life that govern activity inside of cells. Proteins are chains of amino acids, and they take on specific shapes that are deterministic. When one amino acid connects with another, they form a specific shape that helps them carry out their intended actions.
Interestingly, these shapes are deterministic – if we know which amino acid is present and what it is connected to, we can accurately know the shape the protein will take. These shapes are not random – they appear to be guided by pure physics and the specific components available. This is vitally important to biological study, and numerous scientists are at work around the globe to further this research.
We have some methods in biological science for determining the shape of proteins, including X-ray crystallography and cryo-electron microscopy (cryo-EM); the latter is revolutionary in and of itself and has produced the most accurate protein shapes to date. Research has been ongoing since the 1950s, but has proven to be arduous and slow through the use of manual techniques. Computers have only recently been utilized as tools that help this research along.
Some readers may remember Folding@home, which was a small program users could install on their computers to help with protein folding research efforts. It was a small app that would work in the background and/or when a computer was idle. As a distributed program that was running on computers around the globe, it allowed the collective power of the internet to aid scientists as they probed this critical topic further.
John Moult, a computational biologist at the University of Maryland in College Park, founded a competition in 1994 called Critical Assessment of Structure Prediction (or CASP) to encourage research entities to put forth their techniques in the name of progress. Since then, groups have worked on protein structures that are not known publicly, and their results are compared to see who has most closely found the correct shape.
Two years ago, Google’s DeepMind outfit saw their AlphaFold program win the competition, generating results that were excitedly received. While it was using an AI-driven approach that was similar to other competitors, its deep learning algorithms would take its findings to help generate a “consensus model” of the protein’s shape. While incredible, this process hit a wall, and was unable to progress further.
John Jumper – the leader at DeepMind – then worked with his team to develop an AI network to improve its predictions. The results have proven astounding to biological science. “In some sense the problem is solved” according to Moult. CASP gives out a score to each group on a 100 point scale; most averaged a 75, while AlphaFold turned in scores near 90.
Even in places where AlphaFold doesn’t quite perform as well, the raw data is still invaluable. Mohammed AlQuraishi, a computational biologist at Columbia University in New York City and a CASP participant, remarks that, ““I think it’s fair to say this will be very disruptive to the protein-structure-prediction field. I suspect many will leave the field as the core problem has arguably been solved,” he says. “It’s a breakthrough of the first order, certainly one of the most significant scientific results of my lifetime.”
The benefits of such research are difficult to fully understand, but are incredibly exciting. Andrei Lupas, an evolutionary biologist at the Max Planck Institute for Developmental Biology in Tübingen, Germany, believes that, “It’s a game changer. This will change medicine. It will change research. It will change bioengineering. It will change everything.” AlphaFold was able to help him determine a protein structure that his lab had struggled with for ten years.
Applications could mean designing our own proteins, better drugs that are created more quickly, and the ability to find solutions to diseases through the creation of new medicines and therapies. In a world where AI could bring about amazing healthcare benefits, AlphaFold’s work could usher in a new era of study and biological understanding. It cracked a 50 year old problem; the possibilities are endless.
It will take some time for this kind of research to be applied, but scientists are eager to continue AlphaFold’s work. Demis Hassabis, DeepMind’s co-founder and chief executive, says that the company plans to make AlphaFold useful so other scientists can employ it. “I do think it’s the most significant thing we’ve done, in terms of real-world impact.”
