What to do with bioprocess data in pharmaceutical development and commercial manufacturing?

Highlights

• Overview, challenges, and solutions for bioprocess data management
• Solutions for bioprocess data visualization & statistical analysis
• Use of soft sensors in bioprocess analysis workflows
• Werum PAS-X Savvy combines data management and analytics in one platform

As a process engineer working in process development or manufacturing science, you are responsible for efficiently developing high-performing bioprocesses, ensure a robust scale-up and make sure that a commercial process remains in state of control. Consistent data management and data analysis play a critical role in the success of your pharmaceutical development and manufacturing goals. The main challenges are:

• Where is the data and how to get it?
• How to aggregate the data and make it available to analysis?
• How to analyse the data?

If you are working in the field of bioprocess development, scale-up process validation, or manufacturing excellence, you might frequently ask yourself: How do I get a complete and reliable overview of the process data with less effort? How to structure the ample amounts of data from different sensors in a single database? How to aggregate data from small and large scale to perform scale-down model qualifications? Which methods shall I use for pharma data analytics and the statistical evaluation of bioprocesses? 

PAS-X Savvy provides a solution to all of these challenges. This article gives a brief overview of bioprocess data management, bioprocess data visualization, and statistics to evaluate bioprocesses. 

Data management

As a process engineer you have to deal with large amounts of sensor data (e.g., pH, temperature, and dissolved oxygen measurements), product quality data (e.g., product concentrations, specific activity, relative potencies), “non-numerical” data such as pictures (e.g., scanned SDS-PAGE data) and many more. For every analysis purpose, you need to manually reorganize the data in a time-consuming process from different data sources. 

The first and often most critical challenge you face is to get the data in one place. The data is usually spread across many locations in a company. Different departments, different devices and data storage systems. Once you get the data, you encounter multiple types of data that require different treatment.

We classify data into two major classes:

1. Time Series Data

• Data recorded over time (i.e. each value has an associated time stamp)
• Can be further classified in:
  • Data that is recorded on-line by a system 
    • Often available in a high timely resolution
    • Often exportable in defined export formats
• Data that is recorded manually 
  • Often in a low timely resolution 
  • Often captured in varying formats

2. Feature Data (F)

• Single point data (i.e. only a value)
• Can be further classified in:
  • Scalar features
    • Physical quantity that is completely described by its magnitude
  • Categorical features
    • Assigning a unit operation to a particular group or nominal category.

Challenges with time series are that for one run they often are recorded by different systems and might occur in different timely resolutions. Additionally, when comparing multiple runs with each other, it’s required to align them to specific events in order to make them comparable (e.g. time of inoculation for fermentation processes or start of elution in chromatography).

You can gain the highest amount of information by combining all types of data. That requires that you organize your data. If done manually this is very time consuming and needs to be done again if new data arrives. Not seldomly, data analysis projects consist of 80% data mining and alignment and only 20% of the actual analysis.

One solution might be that you use 2nd level systems that you already have in place. In the case of a fermentation process you might be able to use the available SCADA software for data alignment. Third party systems (e.g. off-gas analyzers and weight balances) can often be connected to the software. Some tools even provide the possibility to add manually recorded data. When exporting the data make sure that you adjust the timely resolution to a level where you still get relevant information while keeping the datasat size at manageable level. It’s recommended to set up SOPs for the export of such systems, so that you can make sure that the variable naming is consistent and the data format stays the same.

Best-practice database requirements are i) a suitable database model to store all bioprocess relevant data in one common database, ii) the possibility to add meta information to time-series data (e.g. define process phases and events), and iii) management of data pre-processing workflows.

With it’s unique data model, PAS-X Savvy aligns and contextualizes all data from MES, ELN, LIMS, DCS, historians, data lakes and single devices and makes them available in one single platform. Customizable database filters help you to quickly identify relevant batches, unitoperations and data types and create your own datasets.

Having all data in one place, the next challenge is to select the right visualization technique dependent on the data type.

Visualize data

Visualization is the most essential tool to detect processing trends and to investigate deviations. To detect trends fast and report the results, ideally the visualizations are directly created from a common database, without the need for exporting the data or manually manipulating data in spreadsheets – which is prone to handling errors and not recommended. Each data type requires different visualization techniques, like multi-axis overlay plots to analyze processing trends, and box plots/histograms to relate quality and product attributes, are excellent for viewing your data.

Commonly used visualization tools for bioprocesses are:

Soft Sensors for information mining

Soft sensors make the most out of your data and signals that you have you are collecting already. So instead of buying more and more hardware, soft sensors help you to get more information on your system solely using the data you already have.

Summary

One of the most critical questions a process engineer or process development manager has to answer is how to realize data management, visualizations, and must-have statistics. One possibility is to use spreadsheets. Although this might sound like a quick fix, you will run into severe difficulties in the short and long run. First, you will miss the possibility of smart ways to search for batches and filter for the data you really need. The calculations are time-consuming and rarely reproducible by another scientist or for management review. You will spend much time trying to standardize your Excel sheets using formulas and macros, and in the end, they are likely inconsistent. In the long run, you will lose a lot of time copying and pasting data from spreadsheet to spreadsheet.

A second possibility is to manage your data in a central database and connect it with statistical and visualization software. Various data sources with different data models and data formats have to be handled by the database. The data model will be very complex, as in addition to timeseries data and numerical features, you must store additional information such as run names, process phases, etc. In addition, all tools need to be maintained, and you will quickly find yourself in a costly patchwork of many different software components without the possibility of direct database integration.

The ideal solution is to use a centralized solution like PAS-X Savvy that combines data management, visualization, data analysis and reporting in one place. PAS-X Savvy is used to automate your data management, visualization, and statistical analysis for the bioprocess lifecycle for upstream, downstream, and quality data. The world’s top pharma and biotech companies trust in PAS-X Savvy – as a partner for process engineers – in R&D, scale-up, tech transfer, validation and manufacturing in more than 200 installations.