One of the biggest analytics stumbling blocks for biomanufacturers is the need to prepare data in a way that makes it accessible to analytic systems and valuable to end users. Implementing a DataOps approach, a framework governing and standardizing data throughout the biomanufacturing environment, enables near-real-time analysis and batch-to-batch comparisons.

Rajiv Anand, founder and CEO of Quartic.ai, tells GEN that clients implementing a DataOps approach increased yields between 10–40% and improved batch-to-batch consistency. “In one case, pure yield was up 12% and variability was reduced as much as 49%.”

Achieving such results requires preparing data for consumption by algorithms and models in near-real-time rather than direct analysis by humans after the fact. With this approach, variations can be detected and adjusted on-the-fly, thereby improving outcomes. Consequently, the data must be prepared differently so it can be used in interconnected, automated analysis systems and therefore support batch-to-batch comparisons.

“That’s where DataOps comes in…making data available to any of the tools the analyst chooses to use for either investigations or online monitoring of equipment and processes,” Anand says. That requires standardization, using such steps as data cleaning and filtering, time alignment, contextualizing equipment and process models, and stipulating the naming conventions.

Avoid data lakes in manufacturing

One of the big surprises for manufacturers is that the static data repositories (data lakes) that are so effective for business data aren’t well-suited for biomanufacturing. That’s because biomanufacturing data is multifaceted and incorporates the product, the process, the equipment, and their interactions, he explains, as well as discrete events and data recorded manually in logbooks. “There’s a time element involved. It’s not just static data tables.”

Anand, therefore, advises biomanufacturers to think carefully about the application being deployed and the problem they are trying to solve before they begin preparing the data. “For example,” he says, “putting time-series data in a SQL table and expecting it to work [is unrealistic]. It needs to be time-synchronized,” to account for variability while the batch is being made. Likewise, “What you pivot around should depend on the problem you are resolving.”

Understanding and accounting for these and other fine points of biomanufacturing in the DataOps framework requires the active participation of biomanufacturing professionals on the DataOps team, Anand adds. Teams that incorporate biomanufacturing as well as information technology professionals can better fill in any knowledge gaps and therefore craft a better, more robust DataOps framework.

Previous articleModular Bioprocessing Lines Help Biopharma Cut Costs
Next articleFastest Electron Microscope (Attosecond Resolution) Films Interactions of Light and Matter