A new study, published in Nature, outlines an AI-based approach that could help overcome some of the challenges with zinc finger (ZF) domain design and accelerate the development of gene therapies on a large scale.
The tool, called ZFDesign, was developed by researchers at the NYU Grossman School of Medicine and the University of Toronto. Their hope is that the approach could help treat diseases with multiple genetic causes, such as heart disease or diabetes.
Zinc finger editing
Cys2His2 zinc finger (ZF) domains are one of the most abundant protein structures in the human body. These proteins can be engineered to bind to specific target sequences in the genome and used to reprogramme gene expression. This, of course, makes them very useful for a number of therapeutic applications.
The problem is that ZF domains are notoriously difficult to design – the proteins attach to DNA in complex groups and knowing exactly how each domain interacts with neighbouring proteins is extremely challenging. This has meant tools such as CRISPR-Cas9 are much more popular for gene editing applications – but this approach comes with limitations as well.
CRISPR relies on “foreign” bacterial proteins which would trigger a patient’s immune response and lead to dangerous inflammation. The size of the proteins employed in this approach also makes delivery of treatment an issue. In contrast, zinc-finger editing uses proteins already found in the human body, reducing the risk of an immune response. The proteins are also much smaller – only around 170 amino acids in length – which could lead to more flexible gene therapies and easier methods for delivery in humans.
To overcome the challenges with ZF design and facilitate their use in gene editing applications, a team of researchers (led by study senior author Marcus Noyes) screened over 49 billion protein-DNA interactions and developed a deep-learning model to solve ZF design for any given genomic target.
A new approach
The new technology, called ZFDesign, uses AI to model the interactions between DNA and the ZF domains. It also considers the influence of multiple adjacent finger environments.
“Our program can identify the right grouping of zinc fingers for any modification, making this type of gene editing faster than ever before,” said study lead author David Ichikawa, PhD, a former graduate student at NYU Langone Health. The authors also added that their model “consistently produced ZF arrays across a wide range of targets at high efficacy as nucleases, repressors and activators.” Some of these results can be seen in Figure 1.
Thanks to their work, the design of ZFs for any given target is now available at the push of a button. However, study senior author Marcus Noyes cautioned that more research was needed to better control the ZFs since they are not always specific to a single gene and can cause unintended changes in the genetic code. Their next step is to do just that – refine the AI program so build more precise ZF groupings.
