Researchers at the University of Toronto (U of T) and NYU Grossman School of Medicine have developed a new artificial intelligence technology called ZFDesign that can be used to design zinc finger (ZF) proteins to target any stretch of DNA in humans by targeting the genome. , allowing access to gene therapies for a broader range of health conditions. The researchers fed data from billions of interactions between ZF proteins and DNA into a machine learning model, which can generate engineered zinc fingers that bind to the given DNA sequence.
"Designing zinc fingers to bind to specific DNA targets has been an unresolved problem for decades," said Philip M. Kim, PhD, professor at the Donnelly Center for Cellular and Biomolecular Research at Temerty School of Medicine at the U of T, enabling a new generation of in vivo therapies that have proven difficult to develop using CRISPR and other DNA-targeted technologies."
David Ichikawa, PhD, a former graduate student at NYU Langone Health, added, "Our program can identify the correct zinc finger grouping for each modification, making this type of gene editing faster than ever before."
Ichikawa is the lead author of the paper published by the team innatural biotechnology, entitled "A universal deep-learning model for zinc finger design enables reprogramming of transcription factorsThe researchers stated: "In this study, we introduced ZFDesign, a hierarchical attention-based artificial intelligence model trained on full screens of ZF-DNA interactions representing the influence of multiple adjacent finger environments... ZFDesign captures these influences." to provide a general design model for ZF arrays”.
Diseases such as cystic fibrosis, Tay-Sachs disease, and sickle cell anemia are caused by errors in the order of the DNA bases that encode the operating instructions of each human cell. Scientists can correct these errors in some cases using gene editing methods. Other conditions are not caused by an error in the DNA code itself, but by problems with the way DNA is read by the cellular machinery: epigenetics.
A gene usually works together with transcription factors, which tell the cell how much of this protein to make. When this process goes awry, overactive or underactive genes can contribute to diseases such as diabetes, cancer, and neurological disorders. As a result, researchers have been searching for ways to restore normal epigenetic activity. “Programmable regulation of gene expression would offer powerful research tools and enormous therapeutic potential,” the authors state. "Diseases caused by haploinsufficiency, gain-of-function mutations, or incorrect expression of a gene can be directly treated by modifying gene expression."
ZFs are a common class of human proteins that regulate the expression of this gene, a process that rewrites genetic information into RNA and protein molecules. Scientists have long recognized their potential because they naturally bind to DNA, are less likely to elicit an immune response than CRISPR and related technologies, and are small enough to work with clinical delivery methods. IF editing can change and control genes. Among the most common protein structures in the human body, ZFs can direct DNA repair by taking scissors-like enzymes and directing them to remove defective segments of code.
Similarly, ZFs can also bind to transcription factors and deliver them to a region of the gene that needs regulation. By modifying these instructions, genetic engineers can tailor the activity of each gene.
However, one drawback is that it is difficult to design artificial IFs for a specific task. "...the structurally complicated association of ZF domains with DNA made their design challenging," the authors note. Because these proteins bind to DNA in complex groups, researchers must be able to determine, from myriad possible combinations, how each ZF interacts with its neighbor for any desired genetic change. Therefore, for each new DNA target, scientists would have to develop a new protein in a tedious and often unsuccessful workflow. The researchers noted: "While the potential utility of designer IF arrays has long been recognized, their construction remains challenging as appropriate design code has not yet emerged."
This is not for lack of effort, they continued, as various approaches were used to generate ZF libraries and ZF modules to provide layout ZF arrays. Despite this, the team explained, "these approaches require multiple rounds of tedious selection that either produce ZFs with inconsistent activity or the application of preselected modules that often fail when expressed outside of their selected context."
The recently introduced ZFDesign approach solves this problem with a universal template that offers usability comparable to CRISPR and potentially higher DNA specificity. The technology uses artificial intelligence to model and design interactions. "Since half of human TFs use ZFs to target DNA, we concluded that these endogenous ZF domains can be perfectly replaced by modified ZFs without affecting the regulatory function of the protein," the team commented. "This approach presents the engineered ZFs in the exact context in which the ZFs would naturally occur in the starting protein."
“I think this system balances the requirements for zinc fingers and CRISPR,” said Kim, who is also a professor of molecular genetics and computer science at the U of T. “CRISPR is very well established for basic research, but our system has A lot of advantages. , especially for applications in living systems, since zinc fingers are human proteins and would be safer as injectable drugs."
ZFDesign technology can also create many different proteins that do the same thing, providing more opportunities to bring treatments to the clinic. "Our program can identify the correct zinc finger cluster for each modification, making this type of gene editing faster than ever before," Ichikawa said.
The ZF model was developed with a research group at NYU Langone Health led by Marcus Noyes, PhD, assistant professor of biochemistry and molecular pharmacology in the Department of Systems Genetics at the Grossman School of Medicine.
The model is based on data generated by screening billions of potential ZF-DNA interactions in the researchers' laboratories. Noyes' lab has studied zinc fingers for years and has collected data on 49 billion protein-DNA interactions with zinc fingers, performing high-throughput analysis of multiple zinc finger libraries. Their approach combined two layers of data: interactions between individual zinc fingers and DNA, and between each zinc finger and its neighbor. These zinc finger pair interactions affect DNA binding and therefore gene expression.
The machine learning model developed by Kim and his group reflects the data synthesis approach in Noyes' lab. "Our model is hierarchical, so it uses existing data from the phase one assessment, and a subset of your data from phase two, to develop predictions about which zinc fingers will be compatible with others in given contexts," he said. co-author Osama Abdin. , a graduate student in Kim's lab.
The model is based in part on technology that also forms the basis of ChatGPT, a software application developed byopen AIsimulates a human conversation. The model generates amino acid sequences for ZF proteins using large, highly detailed data sets and techniques similar to natural language processing.
The researchers demonstrated the usefulness of the ZF system by reprogramming human transcription factors, ZF proteins that regulate the transcription of DNA to RNA. Working with Donnelly Center professors Tim Hughes, PhD, and Mikko Taipale, PhD, they optimized the DNA-binding targets of various transcription factors and programmed them to activate or repress various genes. "We present a generalizable design methodology that allows seamless replacement of a natural TF DNA-binding domain to direct TF to any target of interest," the authors explained. "These RTFs [reprogrammed transcription factors] can elicit activation and repression activities similar to CRISPR-based tools and establish these proteins as attractive therapies composed entirely of human components."
The clinical application of reprogrammed transcription factors may address diseases caused by haploinsufficiency (deletion or inactivation of a copied gene, as in some types of cancer and connective tissue diseases known as Ehlers-Danlos syndrome) or diseases associated with toxic gene repeats. such as neurodegenerative diseases. diseases such as ALS, Parkinson's and Huntington's disease.
Kim said the system is already generating blueprints for ZF proteins with clinical potential, and both his Toronto and New York teams were amazed at how well it worked. The new system also shows promise for gene editing and other applications where CRISPR is useful, although its impact is likely to be strongest in the area of transcription factor reprogramming, Kim said. The study authors add that the small size of zinc finger tools not only poses a lower immunological risk, but may also offer more flexible gene therapy techniques compared to CRISPR, providing more opportunities to deliver the tools to cells. right at the right times. patients
"By accelerating the design of the zinc finger along with its smaller size, our system paves the way for using these proteins to drive multiple genes simultaneously," Noyes said. "In the future, this approach could help correct diseases that have multiple genetic causes, such as heart disease, obesity, and many cases of autism."
The developers further commented: "ZFDesign represents a significant advance as ZF design for any target is now available at the push of a button and is open to the academic community to study a variety of academic and therapeutic applications of immunogenicity."
Noyes, an assistant professor in NYU Langone's Department of Biochemistry and Molecular Pharmacology, cautioned that while ZFs hold promise, they can be difficult to control. Since they are not always specific to a single gene, some combinations can affect DNA sequences beyond a specific target, resulting in unwanted changes to the genetic code. As a result, Noyes says the team plans to refine its AI program to create more precise IF groupings that trigger only the desired edit. Noyes is also a member of the NYU Langone Institute for Systems Genetics.
The next step is to improve the specificity of the system. "The current model is designed to optimize the engagement of a specific zinc finger protein with its target, but lacks integrated knowledge of interactions with other targets," Kim said. "Optimizing specificity requires modeling these other interactions."
The researchers are on their way to creating a model that offers more specificity, Kim said. "This work in the areas of epigenetics and gene therapy is very exciting and I look forward to seeing what we can achieve with this technology."
Noyes is a co-founder of TBG Therapeutics, which develops methods to engineer ZF and apply them to treat diseases with genetic components.