Chromatogram review enables evaluation of packed-bed chromatography processes. Typically, that method entails paper-based, qualitative comparison of a profile obtained during a run with another from a reference batch. Such a protocol can be time consuming, resource intensive, and susceptible to operator error. Martin D. Jensen (senior engineer at Fujifilm Diosynth Biotechnologies, FDB) delivered an “Ask the Expert†presentation on 20 October 2020 to describe advantages that his company gained by transitioning from paper-based to digital chromatogram review.
Jensen’s Presentation
Chromatogram review serves as an important in-process control because it can identify cases of poor column packing, resin degradation, and equipment malfunction. For instance, excursions of column outlet conductivity can represent channeling and dead-volume formation. Atypical transitions in an impurity’s UV absorbance can indicate changes in its interaction with a resin (e.g., nonspecific binding). Unexpected interactions between products and resins register as abnormal UV absorbance peaks.
During a standard review, an operator transcribes start and stop times for separate phases from a distributed control system (DCS) into a visualization program. The operator generates a plot, prints it, and compares it with a reference chromatogram. Then, the plot is annotated, signed, and added to the corresponding batch production record (BPR).
Jensen explained that such a manual protocol leaves room for several types of mistakes, including procedural problems and variations in assessment. Paper-based protocols also require one page per phase, cycle, and batch. At the FDB site in Hillerød, Denmark, that equates to ~10,000 pages each year, and precious time and resources are wasted in printing profiles, retrieving and circulating BPRs for manufacturing and quality assurance reviews, and scanning and archiving batch records.
The Hillerød facility streamlined operations by adopting the SIMCA and SIMCA-online software suites (Sartorius). The former handles data from reference batches — including column volumes (CVs), UV absorbance, and conductivity — to enable construction of monitoring projects. The latter collects process parameters from the facility’s data historian in real time. Thus, SIMCA-online software can generate live batch profiles that can be overlaid with data from reference batches for easy detection of aberrant parameters. The system also can set x-axes according to CVs, eliminating chromatogram irregularities stemming from changes to flow rate and system holds.
Jensen emphasized the accessibility of generated data. SIMCA-online software can share profiles across departments instantly, facilitating communication. FDB also has integrated the SIMCA suite with Citrix networking solutions to provide manufacturing partners with remote access to batch data, increasing transparency and collaboration.
Some technical challenges accompanied FDB’s digitalization of chromatogram review. To meet quality requirements, the company needed to optimize chromatogram formatting and ensure that all users would experience profiles in the same way. Editing XML file defaults addressed those concerns.
FDB also needed to develop a data-retention protocol. Regulatory guidance requires that manufacturers be able to consult chromatograms as they appeared when a batch was produced. FDB teams established change- management and version-control procedures and optimized their data historian’s data compression settings to ensure chromatogram recall.
Jensen noted that SIMCA-online users might need to tune execution conditions to ensure that the software pulls data from the correct times. For instance, incorrect CV-totalizer zeroing order can leave inadvertent artifacts on a chromatogram.
Although digital chromatogram review enables scientists to evaluate and share process information more quickly and seamlessly than ever before, Jensen observed that review could be expedited further by training a data model to flag chromatograms for operator review when they exceed given thresholds. To realize such a model, FDB is exploring tools that can quantify peak and transition shapes.
Questions and Answers
Has digital chromatogram review detected more process problems than FDB’s legacy system? Significant problems are rare, but reference batch overlay has helped FDB to identify slight differences between normal and atypical conductivity transitions in historical data.
What difficulties arise in training a data model to review chromatograms? Models must be trained on data from successful and failed runs, but FDB has limited data for the latter. One solution is to establish conservative acceptance criteria such that profiles are approved automatically only when they show minimal deviations from the model.