Gene and cell therapy manufacturing processes would benefit from automation and machine learning for real-time quality control, according to the CEO of a company supplying sensors to industrial R&D and manufacturing companies, including those in the emerging field of synthetic biology.
“Debugging a problem in the therapeutics space is very costly and disruptive, because the cause is often a mystery,” explains Sridhar Iyengar, PhD, CEO and founder of Elemental Machines.
He gives the example of a biotech customer who found that a proportion of their products had radically different growth rates. They were using shaker fermenters, and the vibration rate was different because the screw on one of the holders was loose.
“I often tell computer science friends to imagine a computer program with thousands of lines of code, and each time you run it, the operating system has changed slightly, but you’ve no idea what has changed,” he says.
Machine learning/artificial intelligence (AI) and automation can help researchers and manufacturers identify where a problem lies. As Iyengar notes, “If you’re automating in a low cost, efficient, and systematic fashion, you can cross correlate what caused an adverse event.”
Synthetic biology companies have been early adopters of debugging platforms for biological processes. In part, Iyengar points out, because they are working with yeast and bacteria, which behave more predictably due to their simpler genomes. This makes them more suited to an engineering approach.
He compares companies using yeast or bacteria to churn out a single protein to a race car, where cells can be optimized for production at the cost of keeping them alive. In contrast, gene and cell therapies often require working with cells that vary depending on the patient, and where it’s not clear what constitutes a good baseline performance.
“A race car can be tuned to give power at the expense of comfort. But gene therapy is more like a family car—you need comfort and power,” he says.
This makes gene and cell therapies more dependent on a wide variety of biomarkers for measuring baseline performance, and this, in turn, increases the importance of debugging and real-time quality control.