Proteins are fundamental to all cells. We have learnt much about their biology and function, and have produced proteins in the laboratory to use as therapeutics for a range of diseases. Insulin, growth hormone and monoclonal antibodies are good examples. However, to make more effective therapeutics, we need to understand the functions of more proteins and much of this can be discovered through knowing their structure and their shape.
Genome projects have allowed us to predict the primary sequence of many proteins from their gene sequences. In the 1970s, by denaturing and renaturing ribonuclease, Christian Anfinsen demonstrated that its tertiary structure depended solely on its sequence of amino acids. However, only recently has artificial intelligence solved the 50-year-old protein folding problem of predicting protein structure from amino acid sequences derived from these gene sequences alone (see Box 1). As such, in July 2021, it was a significant milestone for biology when the AlphaFold database was released. AlphaFold solved the problem using artificial intelligence methods that mimic the numerous layers found in neural networks to learn from known protein structures, realising powerful, accurate predictions. As a result, we now have structure predictions for most human proteins, as well as those in 20 other species, with huge implications for protein therapeutics.
Your organisation does not have access to this article.
Sign up today to give your students the edge they need to achieve their best grades with subject expertise
Subscribe
