A neural network searches for the most dangerous COVID-19 mutations
The new tool could prevent potentially dangerous variants of the virus from binding to human cells even more actively. If biologists know about possible mutations in advance, it will help fight them more easily. The authors of the new work used a two-step computational procedure to create a model to predict changes in amino acids, the building blocks for all proteins in the body. During the work, the researchers trained a neural network on experimental data about the activity of a new type of coronavirus depending on changes in a single amino acid. They found that they could predict, with about 80% accuracy, whether certain amino acid changes would improve or worsen the ability of SARS-CoV-2 to infect humans and animals.