The challenge of data integrity for bioprocess Raman monitoring
In the case of Bioprocess Raman monitoring, compliance with CFR 21 part 11 can be certainly considered as a challenge. Indeed, a research approach may involve a wide panel of contributors (IT engineers, Raman experts, scientists, operators, chemometricians etc.). Furthermore, a huge quantity of data is collected, including data to be rejected according to spectroscopic and chemometrics criteria. Above all, the quantity of manual data processing to perform the instrument calibration is significant: analyzer parameter setting, analyzer calibration, spectra averaging, normalization/1st derivative, smoothing algorithm, multivariate analysis (including reference measurements alignment with Raman data, PCA and rejection of data, Partial Least Squares (PLS) and chemometrics model building). If all these steps are performed manually, no chance to be CFR 21 part 11 compliant and/or to avoid a project of implementation for each new process!
Though, the good news is that data processing needed for cell culture monitoring may be quite similar from one process to another, if we consider the wide range of applications that a chemometrics approach can address. The journey proposed by ProCellics is to go from manual data processing to automatized data processing, in the sense of a given pre-defined process. In reality, the flexibility offered by CFR 21 part 11 is that the required access record should mean that some steps may include a choice of the operator in terms of data processing, provided this is recorded. The idea is that these choice steps have to be limited and can be justified by ‘crash tests’ of the model which clearly show the bioprocess monitoring space limits versus the acceptable RMSEP.
We will be attending IFPAC conference at the end of the month and will be very interested in discussing Raman monitoring and data integrity with any interested stakeholders. We are indeed always willing to be challenged!