The pressure from regulatory agencies to increase the quality and safety of pharmaceutical products according to the Quality by Design paradigm led to a fast development of Process Analytical Technology (PAT) in the last decade. The development of effective monitoring tools is of particular importance for monoclonal antibody bio-manufacturing. Spectroscopic techniques, and in particular Raman spectroscopy, seem to have a high potential for process information gathering. The development of a rapid and accurate sensor for the quantification of antibody concentration in protein A affinity chromatography would allow to determine the optimal stop point for the antibody loading phase of a continuous capture process, and, consequently, to maximize the costly resin utilization. The first objective of this work was to implement for the first time a flow-through cell for the real-time in-line acquisition of Raman spectra in a chromatographic environment and use it to collect a library of referenced spectra from breakthrough experiments, performed in a process development setting. The spectral datasets were then used to train PLS regression models, aiming to quantify the antibody concentration at the outlet of a chromatographic column, through the explanation of the correlation between the protein levels and the thousands of predictors contained in a Raman spectrum. Attention was paid to the design, optimization and results assessment of a pre-treatment protocol, able to enhance the prediction ability of the chemometric models. The predictions obtained from these models were quite accurate, with a RMSEP around 45-50 mg/L for the low feed concentration runs, containing samples from 0 to 470 mg/L. All the model diagnostics suggest the feasibility of implementing a PLS-Raman soft-sensor for antibody monitoring in the downstream train. Finally, an Extended Kalman Filter based on the combination of the PLS-Raman measurements with a mechanistic adsorption model was tested as a means to increase the robustness of the antibody concentration prediction, showing the potential to reduce the average estimation error to less than 15 mg/L.
La pressione per migliorare la qualità dei prodotti farmaceutici secondo il paradigma del Quality by Design ha portato ad un rapido sviluppo della ricerca nel campo della Process Analytical Technology (PAT). L’implementazione di efficaci strumenti di monitoraggio è di particolare interesse per i processi di produzione di anticorpi monoclonali. Le tecniche spettroscopiche, e in particolare la spettroscopia Raman, sembrano avere un alto potenziale per la raccolta di rilevanti informazioni di processo. Lo sviluppo di un sensore rapido ed accurato per la quantificazione della concentrazione degli anticorpi nella cromatografia d’affinità permetterebbe di determinare il punto ottimale d’arresto della fase di carica e, conseguentemente, di massimizzare lo sfruttamento della costosa resina. Scopo di questa tesi è mostrare la possibilità di implementare una flow-through cell per l’acquisizione in linea e in tempo reale di spettri Raman in un’unità cromatografica. La cella è stata usata per collezionare una libreria di spettri a concentrazione nota, poi utilizzati per creare modelli di regressione PLS. L’obiettivo è quello di quantificare la concentrazione degli anticorpi allo sbocco di una colonna cromatografica, analizzando la correlazione tra il contenuto della proteina e le migliaia di variabili estraibili da uno spettro Raman. L’attenzione è stata focalizzata soprattutto sullo sviluppo, l’ottimizzazione e la verifica di un protocollo per il pre-trattamento degli spettri, con lo scopo d’incrementare il potere predittivo del modello. Le stime ottenute da questi modelli si sono rivelate essere sufficientemente accurate, con un RMSEP di circa 45-50 mg/L per campioni che spaziano tra 0 e 470 mg/L. La valutazione della qualità dei modelli suggerisce la possibilità di implementare un soft-sensor, basato su Raman e PLS, per il monitoraggio degli anticorpi durante i processi di purificazione. Infine, un filtro di Kalman, che combina la stima effettuata dalla regressione PLS-Raman con un modello meccanicistico per l’adsorbimento, è stato testato come potenziale strumento per aumentare la robustezza della quantificazione, riuscendo a portare l’errore al di sotto dei 15 mg/L.
In-line monitoring of antibody loading in protein. A chromatography using Raman spectroscopy
GARBELLINI, SIMONE
2016/2017
Abstract
The pressure from regulatory agencies to increase the quality and safety of pharmaceutical products according to the Quality by Design paradigm led to a fast development of Process Analytical Technology (PAT) in the last decade. The development of effective monitoring tools is of particular importance for monoclonal antibody bio-manufacturing. Spectroscopic techniques, and in particular Raman spectroscopy, seem to have a high potential for process information gathering. The development of a rapid and accurate sensor for the quantification of antibody concentration in protein A affinity chromatography would allow to determine the optimal stop point for the antibody loading phase of a continuous capture process, and, consequently, to maximize the costly resin utilization. The first objective of this work was to implement for the first time a flow-through cell for the real-time in-line acquisition of Raman spectra in a chromatographic environment and use it to collect a library of referenced spectra from breakthrough experiments, performed in a process development setting. The spectral datasets were then used to train PLS regression models, aiming to quantify the antibody concentration at the outlet of a chromatographic column, through the explanation of the correlation between the protein levels and the thousands of predictors contained in a Raman spectrum. Attention was paid to the design, optimization and results assessment of a pre-treatment protocol, able to enhance the prediction ability of the chemometric models. The predictions obtained from these models were quite accurate, with a RMSEP around 45-50 mg/L for the low feed concentration runs, containing samples from 0 to 470 mg/L. All the model diagnostics suggest the feasibility of implementing a PLS-Raman soft-sensor for antibody monitoring in the downstream train. Finally, an Extended Kalman Filter based on the combination of the PLS-Raman measurements with a mechanistic adsorption model was tested as a means to increase the robustness of the antibody concentration prediction, showing the potential to reduce the average estimation error to less than 15 mg/L.| File | Dimensione | Formato | |
|---|---|---|---|
|
2018_04_Garbellini.pdf
non accessibile
Descrizione: Thesis text
Dimensione
4.32 MB
Formato
Adobe PDF
|
4.32 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/139587