 Hi, I'm Stephen Goldrich and this is a short video abstract introducing some recent work we carried out as part of a joint collaboration between University College London and Metamune. The aim of the work was to determine the root cause of a trisophyte bond that was detected at a novel antibody peptide fusion. The research involved the use of high-troup of microbioreactors, advanced process analytics and the application of multivariate data analysis. Just to give you some background into the project, during the research and developing activities of a novel antibody peptide fusion manufactured by Metamune, we identified some product heteronauties which supposedly identified as trisophyte bond. The modification associated with this trisophyte bond altered both the potency and the physical kind of properties of the molecule and therefore required an in-depth investigation. I'm going to pass you on to Nick who will basically talk to you about trisophyte bond quantification. Measuring a small mass difference in a large protein can be challenging. In this case we needed to measure a 32-dolton difference corresponding to a trisophyte bond in an antibody peptide fusion weighing over 150 kilodoltons. To do this we used targeted mass spectrometry. We took protein-apurified samples, denatured and digested, and then monitored peptides containing the modification. Using a triple quadrupole mass spectrometer, we applied mass filters to distinguish dye and trisophyte bonded peptides and in doing so quantified the level of trisophyte variant down to a single percent. So to investigate which of the key process parameters resulted in high and low trisophyte bond levels, we implemented a three-level fractional factorial design of experiment. One of the big benefits of the high triple bioreactive system is the sheer volume of data which is generated in each run. Considering all the online and offline data, including manipulator and control variables, combined with initial conditions on top of that we have product quality. The analysis of this massive data set becomes a daunting multivariate data analysis problem. To address this issue we tackled it using two different approaches. The first was a multilinear regression model that focused solely on the DO inputs and the second was a partially squares model which included all of the available data. Both of these modeling approaches yielded highly valuable insights into the key process parameters which influenced this trisophyte bond and also generated excellent predictions of the trisophyte bond at point of harvest. To further validate the model we selected some process parameters which yielded about high and low trisophyte bond concentrations and we implemented these at a seven-litre scale and the MVD model generated highly comparable results to the experiment predicted trisophyte bond levels. The insights generated from this work enabled the control limits of the key process parameters to be redefined to minimise the trisophyte bond. And if you are interested in understanding the methodology which we used to combine both the online and offline process analytics to evaluate the powder hedgeronades, please have a look at the paper. And to conclude I'd like to thank all of my fellow co-workers, specifically William Holmes and everyone else that was involved in the project from Metamune and UCL. Thanks very much.