The creation of an artificial intelligence network developed by DeepMind, an AI company owned by Google, has resolved one of life science’s biggest mysteries. The program, named AlphaFold, has the ability to determine the structure of a protein from its amino acid sequence, which encompasses 200 million proteins from 1 million different species.
The artificial program – AlphaFold, has revolutionised the field. Laboratory research has been the primary method of discovering protein structures for decades. The study of one protein may last the entire PhD of a scientist. The time-consuming and expensive methods include X-ray crystallography and cryo-electron microscopy, the conventional tools used to solve protein structure issues. Initiatives have been going on for more than 50 years, but accuracy at the atomic level took much work to achieve (Anfinsen, C. B. 1973). The data generated by AlphaFold is open source, all available to the public, impacting thousands of different applications and opening up new opportunities and research not envisioned before. The data also includes the accuracy of the AI’s predictions so researchers can determine its reliability. It is one more efficient and easy-to-use tool that scientists have now. However, it is essential to acknowledge that this is the culmination of years of effort. Machine learning functions mainly on the idea that there are large amounts of data to feed into it. The Protein Data Bank (PDB) is an essential source for that, which came to be from scientists around the world deposition information into it.

Why are Protein Structures Important?
Amino acids make up proteins, which are life’s building blocks, and are involved in most physiological and body functions. Its three dimensional structure determines its function, affecting cell integrity and shape, highlighting the importance of understanding its structure. For example, most pharmaceutical drugs are developed using information about their structure and target specific sites and mechanisms of the cell. A few examples of processes involving proteins: are DNA replication, cell immune response, energy storage and usage (Breda A, Valadares NF, Norberto de Souza O, et al. 2006).

What are its Applications?
This groundbreaking technology allows various applications, such as the research and investigation of neglected diseases through the discovery of drugs. It has been made easier to identify molecules that are able to treat them. An example is the Chagas disease, predominant in the low socioeconomic areas in the Americas, transmitted through the protozoa Trypanosoma cruzi, affecting millions of people. Research is made easier and more accessible, as AlphaFold facilitates the initial process of determining the protein structure, thus possible targets for therapy. The program is home to 19,036 protein models from the parasite, with various labs using this to find better alternatives to the two anti-parasitic drugs that are currently available and also gain a better understanding of targets (RosLucas et al., 2022). Another major problem in the industry is antibiotic resistance, representing high costs and inefficiency in healthcare combined with the emergence of ‘super-bugs’. This means that insufficient antibiotics have been developed due to the inefficiency and costs of the methods. Furthermore, microorganisms are becoming increasingly resistant to antibiotics, as a relatively regular cycle in the co-evolution of pathogens and hosts. However, that means that new tools and methods are needed to combat the emergence of these resistant pathogens. AlphaFold allows the structure of the bacteria to be predicted in less than an hour when it used to take years (Jumper, J., Evans, R., Pritzel, A. et al. 2021). This encourages a more rapid and efficient identification of target and antibiotic synthesis. Another approach would be targeting the resistance mechanism, which is regulated by enzymes. This is currently being done in labs like the Sousa Research group at the University of Colorado Boulder.

What is Next?
There are still more advances and improvements to be made, but the research standstill in the methodology has leapt. One potential issue is that for engineered or designed sequences, meaning non-natural ones, AlphaFold does not predict whether the predicted structure is stable or not. The reason is that the data used to train the AI were naturally stable sequences. Furthermore, breakthroughs like AlphaFold must be coupled with other advancements like modelling, especially in drug development. Therapy and solutions for critical issues like those mentioned in this article require complex multi-stepped procedures, and AlphaFold is only part of the process.
Most importantly, AlphaFold needs to be tailored for each specific case. For example, chemical and physical properties should also be included in assessing the probability and outcome of interactions, not just looking at the shape. Some of DeepMind’s founders are UCL alums. This breakthrough is significant at a global scale, especially for those that are in this field relating to MedTech. Keep an eye out for future advancements!

1. Anfinsen, C. B. Principles that govern the folding of protein chains. Science 181, 223–230
   (1973)
2. Breda A, Valadares NF, Norberto de Souza O, et al. Protein Structure, Modelling and
   Applications. 2006 May 1 [Updated 2007 Sep 14]. In: Gruber A, Durham AM, Huynh C, et
   al., editors. Bioinformatics in Tropical Disease Research: A Practical and Case-Study
   Approach [Internet]. Bethesda (MD): National Center for Biotechnology Information (US);
3. Chapter A06.
4. The Use of AlphaFold for In Silico Exploration of Drug Targets in the Parasite Trypanosoma
   cruzi. A. Ros-Lucas, N. Martinez-Peinado, J. Bastida, J. Gascón and J. AlonsoPadilla.Frontiers in Cellular and Infection Microbiology 2022 Vol. 12
5. Jumper, J., Evans, R., Pritzel, A. et al. Highly accurate protein structure prediction with
   AlphaFold. Nature 596, 583–589 (2021).