Respuesta :
The latent space models developed with variationally auto-encoders offer a way to explore the wealth of information on evolution, fitness, and stability that is stored in protein sequences, making them a good choice to direct efforts in protein engineering.
Fitness:
- Protein sequences are rich sources of knowledge regarding stability, fitness landscapes, and protein evolution. Here, we explore how these attributes might be inferred from sequences using latent space models trained with variationally auto-encoders. We demonstrate that both evolutionary and ancestor relationships between sequences are captured by the low-dimensional latent space representation of sequences, which is generated using the encoder model.
- The latent space format also enables learning the protein fitness landscape in a continuous low dimensional space, in addition to experimental fitness data and Gaussian process regression. The model is also helpful in forecasting the landscapes of protein mutational stability and assessing how stability affects protein evolution. Overall, we show that latent space models learned via variationally auto-encoders offer a framework for exploring the rich information on evolution, fitness, and stability inherent in protein sequences and are thus well-suited to support protein engineering efforts.
Learn more about fitness here brainly.com/question/18267504
#SPJ4