The objective of this thesis was to investigate the use of mid-infrared spectroscopy(MIR) as a PAT tool in bioprocessing. This was achieved through the development of chemometric models from MIR spectroscopic data. Models were applied to both upstream and downstream bioprocess steps to evaluate the potential of MIR as a PAT tool in each scenario.The first study included a preliminary examination of 8 typical components found in a mammalian cell culture medium. A multivariate limit of detection (LOD) analysis was performed to establish the monitoring potential of the instrument for the given application. This initial work identified the components which were unlikely to be accurately detected, such as glutamine, but also highlighted the components that showed promise, such as glucose. A 7-level experimental design was used to develop partial least squares regression (PLS) models for each of the 8 components, with optimal model errors ranging from 6.03% for glucose to 63.06% for glutamine. An external influence investigation into the factors likely to impact model prediction ability was also performed. A statistical analysis on these influences enabled the significance of the effect to be determined. Finally, all investigative work performed in this study was completed using 2 MIR immersion probes; the first, a fixed conduit immersion probe and the second, a flexible fibre-optic immersion probe. This allowed for a comparative analysis of probes and identified the fixed conduit probe as the most suitable for the given application. The second study applied models developed for glucose and lactate, in the first study, to a series of CHO DP12 cell cultures. The objective of this work was to investigate how effective the previously developed models were at predicting glucose and lactate concentrations when applied in-situ, in a bioreactor. This study explored the at-line application of MIR and showed an improvement of 60.27% for glucose and 13.22% for lactate predictions, when applied at-line rather than in-situ. Central to this study was the optimisation of the models used, in an effort to improve their accuracy and robustness. The impact of the size and nature of the calibration set was investigated to identify the optimum calibration set for in-situ and at-line model development. When offline data was used as part of the optimised calibration training set, models consistently performed better than the original partial least squares regression (PLS)models, created solely from synthetic samples. Optimised glucose model results showed an improvement in RMSEP of 37.93% while optimised lactate model results had an improved RMSEP of 61.98%. The third and final study presented here investigated the use of MIR as a qualitative and quantitative tool for total and recombinant protein detection. Exploratory work to establish the instruments capability in distinguishing between differing proteins was initially performed via principal component analysis (PCA). This showed that the instrument used could identify the recombinant protein of interest among a group of 5 other standard proteins. In-process samples, provided by an industrial collaborator, were used to develop PLS models for total and recombinant protein prediction. Results indicated that models performed better for total protein quantification, with the minimum percentage error of prediction, (PEP) 2.39%. The smallest PEP for the recombinant protein was found to be 6.66%. This study was completed with an investigation into the likely impact of 2 detergents on model performance. Due to the common usage of detergents in protein production, for e.g. protein solubilisation or virus inactivation, the effect of these on MIR spectra and consequently models predictions was investigated. This analysis indicated that detergents were likely to impact results, particularly at concentrations at the higher end of the typically used range. This thesis establishes and evaluates the potential of MIR spectroscopy as a PAT tool.It presents 3 studies which highlight development methodologies and outline possible applications, all the while seeking to optimise results obtained. Through systematic,novel investigations this thesis shows that MIR can be used as a PAT tool, but equally,it raises warnings of when the technique or analysis methods may fall short of the desired result.