The right digital strategy can make lab data accessible to machine learning, as automation continues to sweep through the biopharma industry. Credit: Andriy Onufriyenko/ Getty Images
The biopharma industry is rich with opportunities to boost productivity through automation, from computer-led drug discovery to fully validated vaccines rolling off production lines. But these kinds of innovations don’t spring from nowhere — they are underpinned by crucial layers of digital infrastructure in the lab. Tech-conservative firms hesitant to integrate digital tools and solutions such as electronic lab notebooks (ELNs), laboratory inventory management systems (LIMS) or laboratory information systems (LIS) risk being left at a competitive disadvantage when the AI revolution swings into high gear.
“Pharma companies are far behind many industries in terms of digitalization,” says Oliver Hesse, head of biotech data science and digitalization at Bayer Pharmaceuticals in Berkeley, California. “Partly that’s due to the unique challenges we face, but it also comes from being over-focused on avoiding risk or waiting for the proper use case. That’s a trap — you have to take a more holistic view.”
With Hesse’s background in high-throughput screening, lab automation and data science, Bayer recently tapped him to manage a worldwide team charged with updating much of the company’s legacy equipment for the information age. The lessons he has learned about establishing seamless information transfer from process development to manufacturing could pay dividends to pharma firms looking to progress from digital record keeping to full-on automation.
The value of structured data
While dealing with “systems that don’t play nicely together” has been the primary obstacle Hesse has faced, he notes that changing the mindset of researchers reluctant to adopt digital lab platform (DLP) tools such as ELNs and LIMS ranks a close second. To overcome this, Bayer dedicated a team of biotech engineers to work closely with laboratory users to put their needs at the forefront of a custom-built digital platform.
“For the past two years, my focus has been to create an infrastructure that captures all of a user’s data, and to help people understand the value of structured data,” says Mehdi Saghafi, a biotech data engineer at Bayer with 20 years’ experience in process development. “After a lot of hand-holding, planning and strategizing, it's really starting to flower.”
Saghafi explains that digitalization involves more than replacing paper lab notebooks with tablets, or simply dumping results into ever-expanding hard drives or one-dimensional digital tools. In a truly optimized digital lab, “the data no longer resides on a piece of equipment — it’s available at your fingertips.” The main challenge with implementing this vision, he notes, is finding people with the skill and creativity needed to modernize legacy equipment, workflows and databases using application programming interfaces (APIs).
“Every instrument is different, and there’s no manual to tell you what to do,” says Saghafi. “It requires a certain amount of persistence, and many corporations aren’t willing to fund a group to manage the transition.”
These sentiments are echoed by Zareh Zurabyan, head of eLabNext America, a DLP provider offering tools such as ELNs and LIMS from its base in Cambridge, Massachusetts. “It’s not like getting a centrifuge or a flow cytometer,” he says. “A digital solution like an ELN becomes the centrepiece of your daily routine. As well as unlocking research insights, having large-scale data at your fingertips will, at a minimum, influence your business strategy. We always recommend clients set up a committee to define what the digital strategy is from the outset.”
A holistic view
The ever-growing need to reduce time to market is driving pharmaceutical companies to adopt more efficient, data-oriented processing techniques. Central to that goal is managing data so it’s in the right place at the right time to learn from it. According to the Bayer team, taking a step back proved key to bringing disparate components together in an integrated infrastructure.
“Look at the big picture — what is a bioreactor? A vessel with inputs and outputs. Now how do you control that, how do you fit that into a system?” asks Saghafi. “And think about handling the metadata around that bioreactor: things like the batch, the project, the operator. That’s where an ELN becomes crucial.”
A typical process development setup has a hierarchical structure, with supervisory control and data acquisition (SCADA) software sitting at the top directing traffic between programmes such as a data historian and an ELN that acts as a user interface and a central hub for data analysis and process modelling. “There are lots of tools to transfer your lab analytical data, but if you can’t visualize and analyse them all in one place, they become meaningless,” states Saghafi.
Zurabyan notes that eLabNext has open API and software development kits that allow just about any lab to push and pull data between instruments with ease. “It’s a modular system with fully indexed components, which makes it more intuitive to use,” he says. “Once you get used to it, you can keep adding more capabilities through our online marketplace, which features some of the top third-party AI tools in the industry. The idea is to build an innovation ecosystem to optimize research and process development.”
The ability to find simple solutions for users proved key to driving adoption rates at Bayer. “Don’t overcomplicate things — that was a lesson for us,” recalls Hesse. “If you have 200 codes to memorize, it won’t flow with what you’re trying to do.”
The AI revolution is coming
Although the end user may not need to see it, a considerable amount of infrastructure needs to be in place for the digital lab to be successful. For Zurabyan, labs that make this investment will have a much greater chance of success when the next digital revolution hits. “AI is going to come out of nowhere and change everything,” he says. “When we consult with labs, we really focus on data standardization so it’s accessible to machine learning.”
Saghafi likens these proactive efforts to communities paying taxes for new roads. “Sometimes we have to do uncomfortable things, but look — if you’re any good in the lab you already keep a notebook. Spend a bit of time with taxonomy, learn the proper way to capture and reference data with an ELN so that a person who has nothing to do with the lab can analyze it in its proper context.”
Even with an expanded digital arsenal, innovation in the pharmaceutical industry still needs a human touch. “If implementation happens at the level of the end users and you partner with them to give them the right software, they take ownership of it,” says Saghafi. “The digital lab platform becomes the pillar of your innovation, capturing everything — your data, your repeatability, your future.”