summary: Researchers have developed a deep learning model, TIGER, that accurately predicts the on- and off-target activity of RNA-targeting CRISPR tools. This novel approach enables fine-tuning of gene activity in human cells.
TIGER prediction may improve the design of CRISPR treatments by minimizing off-target effects. RNA targeting technology may help treat diseases caused by overexpression of specific genes, and may provide new ways to combat viral infections.
important facts:
- In this study, we present a deep learning model, TIGER, that predicts the on- and off-target activity of RNA-targeting CRISPR tools and enhances the accuracy of gene editing.
- By predicting off-target effects, TIGER allows precise modulation of gene dosage and could be beneficial in treating conditions such as Down’s syndrome, certain schizophrenia, Charcot-Marie-Tooth disease, and certain cancers. There is a nature.
- This study leverages the strengths of machine learning and deep learning in genomics to leverage large datasets from CRISPR screening to improve therapeutic strategies.
sauce: Columbia University
Artificial intelligence can predict on- and off-target activity of CRISPR tools that target RNA, not DNA, according to new research published in nature biotechnology.
The study, by researchers at New York University, Columbia Engineering, and the New York Genome Center, combines deep learning models with CRISPR screens to control human gene expression in different ways. For example, flicking a light switch to turn it off completely, or pressing a switch. By partially lowering the activity using the dimmer knob. These precise gene controls could potentially be used to develop new CRISPR-based therapeutics.
CRISPR is a gene-editing technology that has many uses in biomedicine and beyond, from treating sickle cell anemia to making tastier mustard greens. They often work by targeting DNA using an enzyme called Cas9.
Recently, scientists discovered another type of CRISPR that instead targets RNA using an enzyme called Cas13.
RNA-targeted CRISPR can be used for a wide range of applications, including RNA editing, knockdown of RNA to block expression of specific genes, and high-throughput screening to determine potential drug candidates.
Researchers at New York University and the New York Genome Center have created a platform for RNA-targeted CRISPR screening using Cas13 to better understand RNA regulation and identify the function of noncoding RNAs.
Since RNA is the major genetic material of viruses such as SARS-CoV-2 and influenza, CRISPR targeting RNA is also expected to develop new methods to prevent or treat viral infections. Also, in human cells, one of the first steps in gene expression is the creation of RNA from the DNA in the genome.
An important goal of this study is to maximize the activity of RNA-targeting CRISPR on its intended target RNA and minimize its activity on other RNAs that may have deleterious side effects on the cell. Off-target activity includes both insertion and deletion mutations, as well as mismatches between guide and target RNA.
Previous studies on CRISPR targeting RNA focused only on on-target activity and mismatches. Prediction of off-target activity, especially insertion and deletion mutations, is poorly studied.
Approximately 1 in 5 mutations in the human population are insertions or deletions, so these are important types of potential off-targets to consider in CRISPR design.
“Like CRISPR targeting DNA, such as Cas9, CRISPR targeting RNA, such as Cas13, is expected to have a major impact on molecular biology and biomedical applications in the coming years,” said New York University. says Neville Sanjana, associate professor of biology at Professor of Neuroscience and Physiology at New York University’s Grossman School of Medicine, principal faculty member of the New York Genome Center, and co-senior author of this study.
“Precise guide prediction and off-target identification will be of great value to this emerging field and therapy.”
in their research nature biotechnologySanjana et al. performed a series of pooled CRISPR screens targeting RNA in human cells. They measured the activity of 200,000 guide RNAs targeting essential genes in human cells, including both ‘perfect match’ guide RNAs and off-target mismatches, insertions and deletions.
Sanjana’s lab, in collaboration with machine learning expert David Knowles’ lab, dubbed TIGER (Targeted Inhibition of Gene Expression by Guide RNA Design) trained on data from CRISPR screens. I designed a deep learning model.
Comparing predictions generated by deep learning models and laboratory tests on human cells, TIGER was able to predict both on-target and off-target activity, compared to previous models developed for Cas13 on-target guide design. It provided the first tool for outperforming models and predicting off-target. – Target activity of CRISPR targeting RNA.
“Machine learning and deep learning have shown their strength in genomics because they can take advantage of the huge datasets that can be generated by modern high-throughput experiments.
“Importantly, we were also able to use ‘interpretable machine learning’ to understand why a model would predict a particular guide to work well,” says Computer Science and Systems, Columbia School of Engineering. Knowles, assistant professor of biology and core faculty member at the University of New Caledonia, said. York Genome Center, and co-senior author of this study.
“Our previous work demonstrated how to design Cas13 guides that can knockdown specific RNAs. Using TIGER, we balance on-target knockdown with avoidance of off-target activity We can now design a Cas13 guide,” said Hans Hermann (Harm) Wessels, co-first author of the study and senior fellow at the New York Genome Institute. Mr. Center previously said he was a postdoctoral fellow in Sanjana’s lab.
The researchers also used TIGER’s off-target prediction to precisely measure gene dosage (the amount of a particular gene expressed) by allowing partial inhibition of gene expression in cells with mismatched guides. I have proven that it is adjustable.
This could be useful in diseases such as Down syndrome, certain schizophrenia, Charcot-Marie-Tooth disease (an inherited neurological disorder) where there are too many copies of a gene, and cancers caused by abnormal gene expression. There is a nature. Uncontrolled tumor growth.
“Our deep learning model not only tells us how to design a guide RNA that completely knocks down a transcript, but also allows us to ‘tune’ it. We can have it generate only 70% of the total,” said Andrew Stern, a Ph.D. student at Columbia Engineering and the New York Genome Center, and co-first author of the study.
The researchers envision that combining artificial intelligence with RNA-targeted CRISPR screening could help predict TIGER to avoid unwanted off-target CRISPR activity, further facilitating the development of a new generation of RNA-targeted therapeutics. .
“As we collect larger datasets from CRISPR screens, the opportunities to apply sophisticated machine learning models are increasing rapidly. Next to the Laboratory, we are facilitating this great cross-disciplinary collaboration, and TIGER allows us to predict off-targets and precisely modulate gene dosages, so many of RNA-targeting CRISPRs in biomedicine. Exciting new applications are possible,” Sanjana said.
Other study authors include Alejandro Mendes-Mancilla and Sidney K. Hart of New York University and the New York Genome Center, and Eric J. Kim of Columbia University.
Funding: This work was supported by grants from the National Institutes of Health (DP2HG010099, R01CA218668, R01GM138635), DARPA (D18AP00053), Cancer Institute, and the Simons Autism Research Initiative.
About this AI and CRISPR research news
author: Holly Everts
sauce: Columbia University
contact: Holly Everts – Columbia University
image: Image credited to Neuroscience News
Original research: open access.
“Prediction of on- and off-target activity of CRISPR-Cas13d guide RNA using deep learning,” Neville Sanjana et al. nature biotechnology
overview
Prediction of on-target and off-target activity of CRISPR-Cas13d guide RNA using deep learning
Transcriptome engineering applications in living cells using RNA-targeting CRISPR effectors rely on accurate prediction of on-target activity and off-target escape.
We have about 200,000 designed and tested here RFXCas13d guides RNAs targeting essential genes in human cells with systematically designed mismatches, insertions and deletions (indels).
We found that mismatches and indels have position- and context-dependent effects on Cas13d activity, and that mismatches that cause GU wobble pairing are more permissive than other single-base mismatches.
Using this large dataset, we train a convolutional neural network (referred to as gRNA-designed targeted inhibition of gene expression (TIGER)) to predict efficacy from guide sequences and context. TIGER outperforms existing models in predicting on-target and off-target activity in our datasets and public datasets.
We show that TIGER scoring combined with specific mismatches provides the first general framework for regulating transcriptional expression and that RNA-targeted CRISPR can be used to precisely control gene dosage. show.
