Researchers at Seoul National University College of Engineering and Yonsei University School of Medicine have developed artificial intelligence (AI) that can improve the accuracy of CRISPR-Cpf1, a gene editing technology.
CRISPR is a vital tool in gene correction technology capable of cutting strands of DNA and selectively correct specific genetic information in cells. It consists of a cleavage enzyme that cuts the DNA and a Guide RNA that is a carrier and a guide that leads to the DNA base sequence targeted by the cleavage enzyme.
It is essential to attach the selected gene scissors to the target DNA sequence to improve the gene correction effect. However, choosing which Guide RNA can precisely target DNA sequences to achieve a sufficient genetic corrective result has been a significant concern for gene researchers all over the world.
|Professor Kim Hyung-bum and Professor Yoon Sung-ro|
The research team led by Professor Kim Hyung-bum of the Yonsei University School of Medicine and Professor Yoon Sung-ro of the Seoul National University College of Engineering developed artificial intelligence that collected the data of the gene scissor efficiency measurement and applied it to a deep-learning model.
“There currently is a computer simulation program that predicts the effect of gene scissors,” Kim said. “However, since the amount of stored information on various types of gene scissors is small, it has not been utilized due to inaccurate prediction values.”
As a result, many researchers have made various kinds of genetic scissors directly and tested each one of the through experiments, Kim added.
As the first step in constructing an AI gene scissors prediction model, Kim released a genetic correction effect information with 15,000 different Guide RNAs, which he obtained by advanced analysis techniques by measuring the activity of gene scissors. Afterwards, Yoon introduced this information to the AI deep-processing technology to present the gene clippers that can achieve optimal gene correction rates under various conditions.
The AI also determines not only the gene sequence of the target site but also whether the gene scissors are structurally well accessible to the target site. Such process makes it possible to predict a highly efficient target part based on the data learned by self-learning.
The correlation between the experimental result and the predicted value of this AI was 0.87, while the correlation of the existing gene scissors simulation program was 0.5 to 0.6. The closer the correlation value is to 1, the higher the accuracy and reliability.
“Through the artificial intelligence self-learning, researchers can obtain the information of the most optimal gene scissor by only making and testing a few gene scissors models,” Yoon said. “This will greatly reduce the time, effort and budget put into developing the correct gene scissors model.”
In the future, as the AI further learns more information on the effect of gene scissors, it will be possible to construct a program that can further improve the accuracy and reliability of the gene scissors effect, Kim added.
The results of the study were published in Nature Biotechnology, a biotechnology journal.
<© Korea Biomedical Review, All rights reserved.>