Bbc Inside Science

Proteomes, the sequences of protein within the DNA of every living thing, are notoriously difficult to model. The usual chemical methods can take months, but a new computational model using the ability of artificial intelligence to learn the complex sequences is able to predict structures within a matter of hours. Thousands of protein structure predictions are now available on a public database for anyone to access. Understanding such proteins is seen as key to treating nearly all disease. It also hold the potential for improvements in fields as diverse as increasing crop resilience to climate change and biodegrading plastic on an industrial scale. Marnie Chesterton speaks with Demis Hassabis from Deepmind which developed the protein structure prediction system, and Janet Thornton from the European Bioinformatics Institute which holds the database. Genetic engineering, the promise and perils, is the subject of a new series with Matthew Cobb on Radio 4, called Genetic Dreams, Genetic Nightmares. He tells u