Artificial intelligence (AI) will change biopharmaceutical discovery and manufacturing according to experts who say these data-rich processes are ideal for the application of computer-based learning.
Drug industry use of AI is expected to increase significantly in all areas over the next few years. A recent report suggested the pharma AI market will be worth $10bn by 2024, roughly 40% larger than its value in 2017 .
Areas like medical imaging, diagnostics and genomics will remain the largest areas of AI application, however, such approaches will be increasingly used in drug discovery according to the report.
This view is supported by research conducted by BiopharmaTrend , which predicted that discovery and target identification will be among the busiest areas of pharma AI activity in the coming decade.
Identifying compounds or combinations of compounds that have the potential to be turned into disease treatments is a laborious and complex process.
Most pharmaceutical company drug discovery efforts are based on screening vast libraries of compounds against potential disease targets and checking for interaction. The process is essentially hit and miss.
In recent years, pharmaceutical companies have sought to improve the odds of finding promising candidates by partnering with AI experts says Dan Faggella, CEO of market research firm, Techmergence.
Faggella cited Pfizer’s collaboration with IBM Watson, which is focused on immune-oncology research, as well as Sanofi’s small molecule drug target identification deal with Exscientia and Roche/Genentech’s partnership with GNS Healthcare as examples.
All the firms involved stressed the potential AI has to help mine existing compound libraries more efficiently and identify connections that may otherwise be missed by researchers as the key drivers for the agreements.
AI in bioprocessing
And AI use is starting to extend beyond the discovery laboratory according to Faggella, who told us “AI-based bioreactor development will probably gather steam in the next two to three years.”
This view was echoed by Wei-Chien Hung, a process development scientist at Alexion Pharmaceuticals.
Hung uses machine learning (ML), a discipline related to but distinct from AI, to improve biopharmaceutical manufacturing processes.
“Nowadays, people usually mix AI and ML but, although they might share some similarity in the initial stage for data manipulation, and final purpose for prediction, the learning process is usually different.
“Machine learning mostly teaches human beings to predict results by feeding them with data, whereas AI is able to look for and select the learning objects itself to perform active training, which is more advanced.”
He explained that, “We use ML to determine the parameters that have the greatest impact on product attributes, like titre and sialic acid content etc.
“We focus on upstream parameters like ammonia, glucose and glutamic acid concentration to find the important variables and to eliminate the variables that have no impact on product quality in prediction.”
There are many potential applications of ML and AI in biopharmaceutical processing according to Hung, who suggested that in the future manufacturers would be using such approaches to predict product quality.
Process control could also benefit from the application of AI according to Hung, who says that the predictive nature of the approach can help prevent problems occurring.
“If certain cell performance is trending high and prediction suggests it is likely to have out-of-spec titers then AI could be used to fine tune supplementation in order to pull the trend back to a pre-defined safe zone.”