By Vivienne Raper, PhD

An early-stage company is using machine learning to design bespoke nonviral delivery methods for future gene therapies. Nanite, set up two years ago, plans to use machine learning to design polymers tailored to gene therapies. The company hopes the polymers can replace viral vectors, as a simpler-to-manufacture way of delivering gene therapies into cells.

“There are many reasons to get away from viral vectors and one is manufacturing considerations,” explains Shashi Murthy, PhD, co-founder and CTO at Nanite. “Viral vectors, by their very nature, are produced in cells and have to be made in bioreactors and, therefore, all the considerations around, for example, manufacturing scaleup, come into play.”

Murthy, who spoke at the BioProcess International Conference in September, hopes that, by using polymers, the company can achieve effective delivery as well as simplifying the biomanufacturing process.

As he explains, the polymers are negatively charged and can be combined with a positive- charged gene therapy with a chemical mixing process.

“They’re designed to self-assemble around their nucleic acid payloads,” he says, “which is very simple compared to a viral vector manufacturing process.”

Machine learning is essential 

According to Murthy, the Nanite process is also simple compared to the manufacturing of lipid nanoparticles, the nonviral delivery method commonly used for mRNA vaccines.

“The state of the art usually involves a two-phase mixing apparatus that combines the payload as well as elements of the lipid nanoparticle,” he says. “And that’s a rather complex mixing process.”

Machine learning is essential to designing the polymers, Murthy emphasizes, because they must be tailored to the specific gene therapy.

“There’s a tight interplay between the payload and in vivo mechanism of action, and the delivery vehicle needed to bring [the therapy] to that location,” he says. “You can’t have one without the other, so they have to be integrated at an early stage.”

The company hopes to use machine learning to explore the polymer design space, giving them a starting point to design a polymer. Subsequently, as the gene therapy moves from the preclinical to clinical stage, they can use artificial intelligence to tweak the polymer to make the therapy work better.

“Ideally, you should end up making delivery vehicles that were completely unanticipated by human intuition,” he says. “And, as you progress through the development pathway, we fully envisage the delivery vehicles being refined or modified considerably, with computational insight into design playing a role.”

Nanite, which was founded by a group of entrepreneurs, is already working with foundations and patient advocacy groups and beginning to build business with pharmaceutical and biotech companies.

Previous articleMachine Learning for Process Validation
Next articleAddressing the Coming GDGs-GLA Shortage