Evaluation and implementation of novel Process Analytical Technologies and advanced data analytics for pilot plant and clinical manufacturing monitor

The aim of this EngD proposal is to generate a data pipeline to analyse highly complex and non-structured data sets recorded across multiple different pilot scale and clinical production facilities and help identify the root-causes of any productivity differences. In addition to evaluating a number of novel process analytical technologies (PATs) in terms of their selectivity, robustness, ease of integration, model development and transferability. A key deliverable of the project will be to streamline the data management, visualisation and analytics of all available USP and DSP data recorded between pilot plant and clinical faculties with the ultimate aim of identifying any bottlenecks within unit operations or process deviations between facilities. The EngD will develop machine learning algorithms (ML) to help quantify the batch-to-batch variability associated within a single unit operation and understand its impact on subsequent unit operations, with a goal of trying to mitigation any productivity changes and/or deviations in product quality. The project will initial focus on available chromatography data. Additionally this project will also evaluate several conventional and next generation PAT devices:


1) A time-gated Raman spectroscopy system (Timegate) that can significantly reduce fluorescence and has the potential to enhance the signal to noise ratio and allow for better quantification of the
bioreactors CPPs and CQAs.
2) A capacitance probe (ABER) that allows for real time quantification of cell membrane allowing for cell viability and density to be controlled in real time.
3) An in-built Raman spectroscopy (Unitive Design) unit that has the ability to be multiplexed thus
allowing for the control of multiple bioreactors in parallel and thus significantly reduces the cost of
implementation of this key analytical device.
4) A newly developed novel sensor technology from Proxisense that has the potential to monitor
compositional changes and density using a simple low-tech sensor based off high precision thermal
conductivity measurements.
5) A new UV sensor from Cloudspec (Marama Labs) a highly versatile device that uses full-spectrum UV-Vis analysis to measure sample with unprecedented detail and could be of benefit to DSP operations.


The PATs will be accessed based on developed metrics associated to cost and performance and the best performing PAT will be will be considered for feedback control utilising the best performing multivariable control schemes identified such as feedforward control loops