Weird science!

Over a decade ago, the application of CRISPR for genome editing was first described as a promising new tool that would enable scientists to make genetic changes with a level of speed and precision that had not been possible before.

In ten short years, not only have scientists been able to unravel new facets of how the genetic code impacts life, but they have also begun developing treatments for genetic diseases that were previously incurable.

But what comes next? Like all scientific advances, the foundational CRISPR-Cas9 tools are not perfect, so evolution is necessary.

They cannot edit every location of a genome, are not highly active in every experimental context and can create serious off-target editing events that pose a very real threat to patient safety.

So, how can we take what nature built as a bacterial immune system and transform it into a set of purpose-built tools to better our world? Enter AI.

Contrary to popular headlines featuring noteworthy generative systems such as ChatGPT and DALL‑E, AI tools are not limited to mimicking human activities.

AI can be leveraged to analyse biological data, including that derived from conventional CRISPR-based experiments, which can identify new drug candidates.

For instance, Recursion Pharmaceuticals has long used AI to empower its scientists to ask questions more easily about its massive repository of biological data.

Its recently announced Lowe model works like ChatGPT but for scientific research while also partnering with chip-maker NVIDIA to build the supercomputer required to enable these AI systems.

Organisations like healx are also leveraging the power of AI to discover new drugs for rare diseases and are emblematic of the wider applications of AI to accelerate drug discovery.

Beyond drug discovery, AI is paving the way to better mine electronic health records in drug trials, improve drug formulations by predicting key attributes such as solubility and accelerate regulatory approval by helping scientists anticipate the optimal pathway through the approval process.

More broadly, firms like Insilico Medicine have developed an AI-powered pipeline spanning target discovery to clinical trials with the aim of reducing the time to bring new medications to market.

AI is even being used to improve genome editing tools themselves. All CRISPR-based gene-editing tools rely on the combination of a guide RNA that targets a specific location in the genome, and a nuclease that performs the actual edit at that location.

Both can be improved using AI. Algorithms can be used to accelerate the process of enzyme evolution, where scientists introduce random mutations to a protein to alter its function.

Coupling AI with directed enzyme evolution experiments in the lab can narrow the size of the screen, making it more practical and affordable, leading to the development of improved CRISPR enzymes.

Language models trained on protein sequences are also able to create entirely new proteins, not just improve existing ones.

They have already been used to develop new gene-editing proteins that are just as active as Cas9 but produce fewer off-target effects.

Armed with large data sets of bacterial genomes, advanced AI models can also tease out which proteins might have the most desirable properties for a particular application well before work begins at the lab bench.

AI models have also been used to improve the guide RNA design process. For example, CRISPR interference (CRISPRi), a genetic perturbation technique, uses a non-cleaving Cas enzyme to bind to a genomic target and interfere with a specific gene’s expression.

AI is able to improve the design of these guides by considering factors related to both the guide RNA and the target gene, improving their ability to predict which guides would be successful.

We use sophisticated protein engineering techniques to improve existing CRISPR systems and have created the Alt-R HiFi Cas9 protein – a high-fidelity mutant of SpCas9 that dramatically reduces unwanted off-target editing.

We’ve also developed AsCas12a and LbCas12a ultra proteins that enable useful genome editing and replacement efficiencies.

While these powerful systems can enable scientists to advance their clinical therapies, these technologies can be labour-intensive to develop.

IDT is working to employ AI-based protein engineering techniques that can dramatically reduce the time to engineer these critical cell and gene therapy innovations.

In the end, it is accelerating this pace of innovation that is critical if we are going to impact patients positively.

Modernisations are enabling scientists to move faster and more confidently from theory into the laboratory and beyond.

AI can enable the discovery of new tools with features that move us closer to treating genetic diseases that had no hope of a cure prior to the CRISPR revolution.

These tools might be safer, faster, more potent, or feature some other improvements, but the point is that these improvements must be deployed rapidly to help those in need today.

Shortening the development timeline for new genomic medicines is crucial and largely hinges on the improvements that AI offers.

AI is changing our world in so many ways. Many of these impacts are seen in our daily lives, but we shouldn’t forget the tremendous impact this technology is already having on accelerating life‑changing research.

We are still in the early days of exploring how this technology will change genome editing, but if these early successes are any indication of what’s to come, we have plenty to look forward to.

Adam Chernick is Senior Commercial Product Manager-Genomic Medicine & CRISPR at Integrated DNA Technologies. eu.idtdna.com