MIT Researchers have developed Equidock, a machine-learning model to predict the synthetic antibody complex that forms when two proteins bind together. Synthetic antibodies can bind with a virus’ spike proteins and prevent it from entering a human cell.
Equidock predicts protein structures closer to actual structures observed during research experiments and works 80–500 times faster compared to state-of-the-art software methods.
“Deep learning is very good at capturing interactions between different proteins that are otherwise difficult for chemists or biologists to write experimentally. Some of these interactions are very complicated, and people haven’t found good ways to express them. This deep-learning model can learn these types of interactions from data,” said Octavian-Eugen Ganea, a postdoc in the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) and co-lead author of the paper.
Equidock focuses on rigid body docking — which occurs when two proteins attach by rotating or translating in 3D space. The 3D structures of the proteins are converted into 3D graphs that can be processed by the neural network. The model is incorporated with geometric knowledge so it understands how rotation of an object affects its orientation in 3D space.
Equidock can predict the protein complex in one to five seconds once the model is trained. However, the creators of the model faced a tough time in training the model due to lack of enough experimental 3D data for proteins.