PEGS Boston: Accelerating Antibody Engineering with Structured Data and Automation

Structured Data

The use of structured data and automation can accelerate antibody engineering as much as 50%, Prem Mohanty, product marketing manager and life sciences product specialist for Benchling, told BioSpace.

Speaking before his presentation at PEGS Boston Virtual, he noted that siloed information systems were common in scientific organizations, where they delay development, promote needless repetition of effort, and risk introducing errors into the system.

There are two key challenges: rapidly advancing science and software limitations.

The advances in science mean researchers need analytics for early risk assessments to determine whether molecules can be developed and the ability to work through big data to optimize processes. They also need to integrate quality by design and design of experiments into each aspect of antibody process development and thus enhance quality. As Mohanty noted, “Monoclonal antibodies (mAbs) once dominated. Now, the next wave of antibodies – bi-specifics and antibody drug conjugates (ADS) – is almost here, requiring new screening methods. Companies are struggling with the tools to link their systems and keep up with advances.”

The second challenge is disparity among the software used by scientists and management. “Antibody engineering is complex and is becoming more so as expectations for new candidates become higher,” he began. Today, “Scientists typically focus on the science itself, adopting solutions that specialize in one aspect of the workflow and, over time, these legacy solutions lead to siloed data systems.” For example, software typically is used for the statistical design of experiments and for results analysis. Process development studies occur independently and use other systems to manage raw data and store processed data, making quality by design efforts more challenging than they have to be. That becomes a problem when scientists in other groups can’t access the data they need easily, on-the-fly.

Sharing information from stand-alone systems requires a lot of manual involvement and in-person meetings to keep everyone updated as data is entered, analyzed, and reported across in silico, in vitro, and in vivo screenings. In contrast, having a broadly accessible single source of truth for all immunogenicity information leads to better insights and better predictive methods throughout R&D activities.

As antibody engineering embraces next-generation practices, it needs to focus on three things: lab automation, flexible modeling, and integrated software throughout the enterprise, Mohanty said.

“Companies are optimizing more and more workflows,” he acknowledged. Automation is going beyond simply adding robots to automating ways to set up and run experiments, and acquiring and analyzing data.

Modeling needs the flexibility to handle high throughput applications. Benchling’s software, for example, has high throughput bulk functions that, essentially, multiplexes antibody analysis. Rather than analyze one plasmid at a time, the software can analyze hundreds of plasmids simultaneously. This is true for sequence analysis as well as for the other samples, such as cell lines, that are used to produce the final antibody.

“Flexible modeling allows relationships to evolve as new entities are added to the mix. Having the software to model this is common, but it generally is siloed,” he explained. The sequence analysis team may use one tool – Geneious or SnapGene for genome analysis, for example – and then hand its work to the expression team, which uses different software.” The result is information that is fragmented across its research teams.

“The key to making an ecosystem-wide data access work is having a structured data table,” he continued. “Structures data tables force researchers to capture data in a structured manner…so it can be understood across the platform. When data is captured this way, there’s no cleaning needed.”

Initially, however, as data migrates from legacy systems, data normalization is vital to ensure the researchers use the same terms and that the system recognizes those terms, abbreviations, and even misspellings. To do this, “We can import spreadsheets into our smart tables, and the scientist can see ‘this is the name we use’ and put the data in context,” he explained.

Benchling’s platform was designed in concert with scientists. “There are three layers,” Mohanty said. “We automate the sequence design process. For example, the software can detect the sequence and prepopulate the data fields or other manual processes. Then we can unify the software ecosystem to ensure the data is accessible across all applications. As a result, a notebook entry automatically shows up in the data warehouse and across the entire ecosystem. Finally, we can unify multiple different analytical instruments and bring their data in Benchling, to reduce the chance of entry errors,” Mohanty explained.

Deploying a data management system designed specifically for researchers that is integrated across the enterprise and sits atop scientists’ usual applications dissolves the data silos and enables wider access to scientific teams throughout the organization. Consequently, he said, researchers are working off the most current data, manual entry (and entry errors) are minimized, and the research organization has a single source of truth.

Back to news