A group of Russian bioinformatics have proposed a new neural network architecture capable of evaluating how well a guide RNA has been chosen for a gene editing experiment.
Their approach will facilitate more efficient DNA modification with the popular CRISPR / Cas method and thus help to develop new strategies. to create genetically modified organisms and find ways to treat serious inherited disorders.
CRISPR / Cas
Skoltech researchers led by Konstantin Severinov have used the deep learning, Gaussian processes, and other methods to make the selection of optimal guide RNAs more precise.
The team produced a set of neural networks, that is, trainable mathematical models implemented as sequential matrix multiplication: large sets of numbers with a complex internal structure. A neural network can learn because it has “memory” in the form of numbers that alter in a particular way each time the system completes a calculation in training mode.
The team trained the models on different data sets containing tens of thousands of experimentally validated guide RNAs that had shown high precision performance in human and animal cells.
The researchers proposed an algorithm that estimates the probability of DNA cleavage for a given guide RNA. The resulting scores can drive the experimental design for any CRISPR / Cas-based application. The team used their neural networks to create a set of guide RNAs to make precise changes to the genes on human chromosome 22.. This has been made possible by the high precision of the division frequency prediction and a prediction uncertainty estimation function, which none of the previously existing methods provided.
The findings can be used for a variety of CRISPR / Cas-based technology applications, such as the treatment of genetic disorders, agricultural technologies, and basic research experiments.