The machine learning algorithm boosts the protein engineering process

Oct, 2021 - By SMI

There are numerous machine learning algorithms, only a handful of them take into account the target protein's evolutionary history. This is where the deep-learning algorithm ECNet is used.

Proteins are the molecular machinery of all live cells, and they have been used for a variety of purposes, such as medicines and industrial catalysts. Protein engineering is used to increase protein properties such as stability and functioning in order to overcome the limits of naturally present proteins. Machine learning algorithms help in protein engineering by lowering the experimental load of approaches like directed evolution, which requires numerous rounds of mutagenesis and high-throughput screening. They function by stimulating and forecasting the efficiency of all potential target protein sequences after being trained on protein sequence databases.

Researchers were able to use ECNet to examine the target protein and all of its homologs to determine which residues are linked together and as such crucial for that specific protein. The data is then combined and the deep learning system is used to determine what types of mutations are crucial for the target protein's function. The researchers demonstrated that ECNet outperformed existing techniques on three deep mutagenesis datasets in a benchmark analysis. In a subsequent study, ECNet was used to design TEM-1 β-lactamase—an enzyme that imparts resistance to lactam antibiotics and discovers variations with better fitness and hence resistance to ampicillin.

Moreover, throughout the study, ECNet favored higher-order and new mutations. Using a computational system that can correctly anticipate higher-order interactions helps minimize the amount of time spent on experiments. To enhance prediction efficiency, researchers are integrating all of the proteins in the database with the target protein's particular evolutionary history. They can then utilize the mutations generated by their research to enhance and train the model further. This model is still in development, but it is an upgrade over what is previously known in the research. ECNet is already being used by researchers to build enzyme catalysts with higher selectivities.