AI will help genetic engineers
A group of researchers from the College of Engineering at Seoul National University and the Medical School of Yonsei University have created artificial intelligence (AI) that improves the accuracy of CRISPR-Cpf1 genome editing technology.
CRISPR is an important tool of gene editing technology, which allows you to cut out certain sections of DNA and selectively correct the genetic information of cells. It consists of an enzyme that separates DNA and a guide RNA that "indicates" to the enzyme the DNA section that needs to be split off.
In the process of genome editing, it is very important to precisely aim the genetic scissors at the DNA target. This is the main problem of CRISPR technology: it is quite difficult to choose the right guide RNA that can accurately indicate the edited DNA site to the enzyme.
A group of researchers developed an AI that collected data to evaluate the effectiveness of genetic scissors, analyzed them and created a training model.
A computer simulation program predicting the effect of genetic scissors already exists. But, unlike the new AI, the amount of information collected is not enough to generate an accurate prediction of the result.
The first step in creating a model for predicting the accuracy of genetic scissors was to collect information about 15,000 different guide RNAs obtained through the use of advanced analysis methods. All this information was introduced into AI to determine those "scissors" that will help to carry out the most accurate gene editing. In addition, the AI determined the availability of certain RNAs for the target DNA sites. This process allows you to identify those genes that can be edited with maximum accuracy.
The correlation between the result obtained during the experiment and the prediction determined by the AI was 0.87. For comparison, the correlation of the experimental result with the forecast of already existing programs was 0.5-0.6. The closer the result is to 1, the higher the accuracy and reliability of the forecast.
AI is capable of self-learning: researchers can get a prediction of the result of editing a gene of interest by testing several models of gene scissors before doing so. This will significantly reduce the cost of time, effort and material resources.
As information about different gene scissors accumulates, the accuracy and reliability of AI predictions will increase.
Article by Hui Kwon Kim et al. Deep learning improves prediction of CRISPR–Cpf1 guide RNA activity published in the journal Nature.
Aminat Adzhieva, portal "Eternal Youth" http://vechnayamolodost.ru .