How Life Science Companies Can Harness the Predictive Value of Data to Drive Deeper Process Understanding

March 28, 2022

Contributed Commentary by Claire Hill, IDBS

March 28, 2022 | As the biopharmaceutical industry begins to embrace advanced data collection and analysis tools, it finds itself facing a challenging question: How can drug manufacturers best harness all that data to improve capacity, reduce bottlenecks and cut costs? Establishing a strong digital backbone is the first critical step. From there, manufacturers need to figure out how to strengthen this digital backbone with technology that not only makes the data easily accessible, but also empowers manufacturers to analyze it and use it to make transformative predictions.

To get there, biopharma companies will need to embrace artificial intelligence (AI), and that will require changing fundamental work processes. As described by the BioPhorum Digital Plant Maturity Model, digital transformation requires a connected infrastructure where teams can collaborate around desired outcomes. Moving away from legacy data administration tools like Excel and standalone software is essential—but it’s only the first step. Manufacturers must consider how emerging techniques such as in silico modeling to predict the 3D structure of proteins can also predict how complex biological products will perform under various manufacturing conditions.

That's a massive, revolutionary change in the way biopharma manufacturers work currently. To put this in perspective, gathering descriptive and diagnostic data to understand what happened during a manufacturing run and why it happened is still highly manual, usually involving paper records and spreadsheets. But the potential to help drive faster insights, which in turn will improve efficiency and yields throughout the production process and accelerate tech transfer, is a strong motivation. The ability to capture data across multiple processes and devices—and analyze it in real time—will allow manufacturers to improve process development, streamline tech transfer and obtain a better understanding of the robustness and scalability of each new drug hitting the production line.

Mechanistic modeling is one example. Biological processes are extraordinarily complex, but certain physiochemical properties, such as how strongly a molecule binds to a surface under different conditions, can be predicted using mathematical representations. Instead of running endless experiments in wet labs to perfect purification processes such as chromatography and virus inactivation, much of that can be modeled with AI instead. The caveat is that these models are not 100% accurate, but they can be continually refined by combining data from workflow inputs with data simultaneously drawn from chromatography systems and other equipment. Combined with other approaches such as high-throughput process development (HTPD) that data can then be used to model the various steps of the purification process, establish critical process parameters, and ultimately speed up process development.

Consider another example: A scientist observes poor product quality in the initial production runs of a new drug, with a notable decline in quality happening on day 10. He suggests adding a bolus feed of an essential amino acid on day 7 to fix the issue. The manufacturer can test different quantities of bolus across multiple bioreactors, using near real-time data collection and analysis to determine the effect of each change on productivity and quality. As data flows onto computer dashboards, the manufacturer can perform in-depth analyses across multiple angles and scales to predict how the quality of the end product will change depending on how and when the bolus is added. Where possible, the scientist can also draw on lessons learned from relevant full-scale manufacturing runs.

And the capabilities AI offers to drug manufacturers will only build over time. For example, it’s possible today to predict the folding of a protein using resources like the new AlphaFold Protein Structure Database, from DeepMind and the European Molecular Biology Laboratory. This is an incredible achievement for AI, and advances in adjacent areas such as predicting glycosylation behavior and mechanistic understanding of antibody interfaces can also improve the development of complex therapeutics such as bispecific antibodies.

Predictive modeling won’t completely replace lab testing. But if manufacturers can reduce the number of tests they have to run and variables they have to check, they could cut months off their development programs, because they’ll be able to home in on the factors that are really going to influence product quality and yield. With sophisticated data analysis, manufacturers can quickly identify the problems they need to immediately address—faulty processes or raw materials that reduce yields on a new product by, say, 10%—and devote the necessary resources to fixing those issues.

The holy grail of manufacturing intelligence is the ability to predict problems before they occur. As data-analysis tools continue to evolve, they promise to arm manufacturing and process development teams with the tools they need to identify potential problems and correct them before the scale-up of the production process. It could be as simple as designing a simulation on the computer, pressing a button and then coming back a few minutes later to find precise instructions for adjusting the manufacturing process. When applied across the biopharma lifecycle that won’t just shorten timelines for development efforts, it could also cut millions of dollars off the final cost.

We believe that is the future of biopharma development, though there are a lot of missing pieces that need to be in place before we can realize that vision. Many manufacturers still need to move away from paper-based data collection. Data needs to be digitalized, standardized and moved out of silos. Once the digital backbone and supporting technologies are in place, the opportunities to use AI to improve the development of tomorrow’s cures are endless.


Claire Hill leads IDBS’ market engagement for BioPharma Lifecycle Management. With over 15 years of business analysis and consulting experience, Claire has in-depth knowledge of the data management challenges in the biopharmaceutical industry. Claire has an MSc in Biochemical Engineering and an MBA and previously worked as an analyst for IBM and a consultant for BioPharm Services.  She can be reached at