SCIENTIFIC EDUCATIONAL CENTER science idea

Protein engineering is a branch of biotechnology that deals with the search and creation of proteins with improved or completely new functions. One of the main experimental methods of protein engineering is directed evolution, in which a protein of interest to scientists passes through several cycles of mutagenesis and subsequent selection of successful variants. While this approach is effective, it does not test all possible sequences and requires significant experimental effort.
Machine learning methods are helping to broaden the search for new proteins and to better understand the sequence-function relationship. In the new study, US scientists used deep learning to take evolutionary context into account when predicting protein function. Details about the work of the ECNet algorithm (English evolutionary context-integrated neural network - a neural network integrated into an evolutionary context) are described in an article published in the journal Nature Communications.
To predict the function, ECNet uses information about the amino acid sequence of the known "relatives" of the protein under study - so the algorithm learns which amino acids are functionally linked and are important for the protein to work. The new approach turned out to be more accurate than other existing methods for predicting protein function based on machine learning.
To experimentally confirm the effectiveness of the algorithm, ECNet was used to create β-lactamase TEM-1, an enzyme that confers resistance to β-lactam antibiotics, and to identify variants that increased resistance to ampicillin.
Researchers are now using ECNet to develop enzyme catalysts with improved selectivity.
Article published in the journal Nature Communications
PHOTO © depositphotos
Source: naked-science.ru, sci-dig.ru