A report in Trends in Biotechnology, on the potential for “mechanistically oriented in silico processes” in vaccine development, supports the idea that, whenever possible, developers can save time and cost by running virtual experiments on their computers.

In this paper lead author Christos Varsakelis, PhD, a scientist at Janssen Pharmaceutical, describes fluid mechanics, thermodynamics and transport phenomena, intracellular modeling, hybrid modeling and data science, and model-based design of experiments as “pivotal competences.”

Varsakelis is talking about “digital twins,” an interesting if old idea based on computer simulation. This concept, which first arose around five years ago, has become overused. “And here is the caveat,” he says. “The definition of digital twin has deteriorated to the extent that even small components of the original concept are nowadays being labeled as such.”

As they say everybody wants to get into the act, and this particular act is based on the fact that digital twins are “at the heart” of Biopharma 4.0, according to Varsakelis. So while a fair number of companies are deploying various components, few have taken the full plunge.

“If we confine ourselves to fully functional digital twins then we are at the bottom of the adoption curve,” he explains. “Overall, we could say that while biopharma has begun to entertain the idea of digital twins and their added value, they are only making the first contact with the challenges that have to be faced for their development.”

However developers implement 4.0 or digital twins, they will probably not be able to follow a pure artificial intelligence approach to solving bioprocess issues, Varsakelis tells GEN.

“A big problem that we see is the desire of (bio)pharma to mimic Google with respect to utilization of AI. This is irrational because in bioprocessing we seldom have enough data to adopt a pure AI-driven approach,” he says. “And unlike the human-based questions that Google is trying to tackle, the fundamental laws of nature, as they relate to bioprocessing, are known. This explains why a hybrid approach, where classical process models are enhanced with AI, has a much higher probability of success at a fraction of time.”

Which processes will gain most, or first, from this marriage of AI and conventional process development? After considering amenability to in silico modeling and business impact, Varsakelis taps chromatography as the “lowest hanging fruit” because it is “an expensive step, but at the same time the process is well understood. Bioreactors are also amenable to in silico descriptions but doing so requires a non-negligible R&D phase.”

The goal of next-generation process development, whether it is performed in silicon or in microbioreactors, is to trim development times. Given the months-long vaccine development timelines we’ve recently seen, how things could get any better (or faster) than that?

“It’s been shown that in silico process development can reduce development times by half; an improvement which, at least in principle, should apply for all biologics,” notes Varsekelis. “On the other hand, the development of COVID vaccines was somewhat sui generis, especially with respect to clinical trials, priority for production, and potential markets. But [with other conditions remaining the same], in silico development can still reduce CMC (chemistry manufacturing and controls) development timelines and lead to optimized production processes, and should remain invariant for every development campaign utilizing traditional process development.”

Previous articleWorking Back Towards Robust Cell Therapy Manufacturing
Next articleIndustry Warming to 24/7 Production