Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Elution Trajectory Optimization for Chromatographic Separation

Wacker, Philipp LU (2025) KETM05 20251
Chemical Engineering (M.Sc.Eng.)
Abstract
In this thesis, the traditional way of conducting chromatographic separation by linear elution was re-evaluated by introducing variable elution buffer trajectories in the mobile phase. Through numerical simulation of ion exchange chromatography using a finite volume method approach, various salt concentration profiles were optimized for improving the separation of components, using yield and productivity as the performance metrics. In three different case studies, the procedures were numerically trialed on peptide, viral genome, and protein mixtures. The protein case results were also verified in a lab-scale experiment, utilizing optimized model control parameters.

Optimization of the buffer trajectory was studied using two approaches,... (More)
In this thesis, the traditional way of conducting chromatographic separation by linear elution was re-evaluated by introducing variable elution buffer trajectories in the mobile phase. Through numerical simulation of ion exchange chromatography using a finite volume method approach, various salt concentration profiles were optimized for improving the separation of components, using yield and productivity as the performance metrics. In three different case studies, the procedures were numerically trialed on peptide, viral genome, and protein mixtures. The protein case results were also verified in a lab-scale experiment, utilizing optimized model control parameters.

Optimization of the buffer trajectory was studied using two approaches, namely single- and multiple shooting optimization, where single-shooting became the main method applied in this thesis. Due to a high co-variance between variables and the requirement of multiple levels of optimization, the objective functions became noisy and non-convex. This was solved by adjusting the objective of the pooling algorithm and by introducing genetic algorithms, where Differential Evolution was used for singular objectives and the Non-Sorting Genetic Algorithm was used for multiple objective optimization.

The optimization results varied significantly in different cases, but the step elution (i.e., zero order hold) methods were generally deemed to outperform other hold orders. At higher trajectory degrees of freedom, the optimization convergence worsened significantly. The obtained trajectory solutions showed large performance improvements in the viral vector case, while resulting in minor improvements in the protein case and intermediate enhancement in the peptide case. There was generally no increase in optimal yield, but large increases in performance were often found from the productivity perspective. However, when including the column productivity in the objective, only the first couple of segments were proven to influence the result. This indicates that the total elution time is a critical factor as the end of the elution window often contained a lot of 'dead' time with no performance impact. It was also shown that optimizing an additional flow rate trajectory yielded little improvement compared to only applying a buffer trajectory at a constant flow rate.

The experimentally trialed protein case was calibrated to the numerical simulation model and showed a large improvement in peak separation. Although the model predicted little performance improvement between a linear gradient and the segmented variable profile, the practical experiment was deemed to perform significantly better, due to an improved peak separation at a similar peak retention time for the target component. Thus, the trajectory optimizations were deemed to be a success. Their implementation in a lab-scale chromatography system was shown to be straight-forward, and could likely lead to an increased production efficiency in many other applications, enhancing the downstream process of biopharmaceuticals. (Less)
Popular Abstract (Swedish)
Biologiska läkemedel används i allt större utsträckning. Med tiden upptäcks fler och fler användningsområden för exempelvis proteiner (i form av hormoner, antikroppar och enzymer) eller virus, som kan användas till bland annat cancerterapi, genterapi eller behandling av neurologiska sjukdomar. På senare tid har marknaden exploderat genom bland annat upptäckten av GLP-1-analoger som har enorm potential för diabetes- och fetmabehandlingar. Biologiska läkemedel produceras ofta genom att odla en cellkultur, där djurceller genetiskt modifieras för att producera det önskade läkemedlet, varefter man extraherar läkemedlet ur cellerna. Det är oftast genom just detta produktionssättet som man klassifierar ett läkemedel som biologiskt.

Problemen... (More)
Biologiska läkemedel används i allt större utsträckning. Med tiden upptäcks fler och fler användningsområden för exempelvis proteiner (i form av hormoner, antikroppar och enzymer) eller virus, som kan användas till bland annat cancerterapi, genterapi eller behandling av neurologiska sjukdomar. På senare tid har marknaden exploderat genom bland annat upptäckten av GLP-1-analoger som har enorm potential för diabetes- och fetmabehandlingar. Biologiska läkemedel produceras ofta genom att odla en cellkultur, där djurceller genetiskt modifieras för att producera det önskade läkemedlet, varefter man extraherar läkemedlet ur cellerna. Det är oftast genom just detta produktionssättet som man klassifierar ett läkemedel som biologiskt.

Problemen med biologiska läkemedel är att de ofta är väldigt dyra. Produktionen är komplicerad, och ofta så krävs det många år av forskning och utveckling för att testa och upptäcka de önskvärda resultaten och minimalisera bieffekterna. En av de största utmaningarna är uppreningen, där uppgiften är att separera en av miljontals molekyler som finns i en animalisk djurcell. För att separera ut dessa önskvärda ämnen används en rad olika metoder, som är baserade bland annat storlekskillnader och laddningskillnader. Ett exempel på en separationsprocess är jonbyteskromatografi, där laddningskillnaden på ämnena används för att separera dem.

Jonbyteskromatografi innebär att man låter en laddad komponent fastna på en laddad yta, och sedan ökar saltkoncentrationen fram tills att produkten tvingas lämna ytan och rinner ut. Inom traditionell jonbyteskromatografi används ofta en linjär saltökning, där idén är att ämnena som är mer och mindre laddade rinner (eluerar) ut strax efter och innan produkten, så att man kan samla in (fraktionera och poola) produkten i det tidsfönstret där den rinner ut. Detta examensarbete utforskar möjligheten att använda tidigare beprövade modeller för proceduren för att matematiskt optimera den optimala salt-trajektorien under elueringsfasen, och på så vis nå en bättre prestanda för processen. Trajektorien behöver då alltså inte vara en linjär gradient, utan kan anta vilken form som helst under några randvilkor. Prestandan mäts i form av olika mål i renhet, utbyte och produktivitet, som är tre viktiga mått inom läkemedelsindustrin. Utöver detta utforskades även möjligheten att variera volymsflödet under elueringen för att förbättra prestandan, alltså att optimera två olika trajektorier samtidigt.

I denna avhandling studeras simuleringar för peptid-, virus- och proteinlösningar, där den sistnämnda även beprövades i en experimentell fallstudie. De nya variabla profilerna visade olika förbättring i prestanda i olika målfunktioner, och resultaten varierade kraftigt beroende på vilka komponenter som studerades. En stor del av arbetet handlade om att det var mycket svårt att fullborda optimeringen, delvis eftersom problemet behövde optimeras på flera nivåer. Resultaten visar även på att de nya salt-trajektorierna är relativt lätta att integrera i befintliga, verkliga system. Dessutom stämmer resultaten överens efter att ha implementerats på labbskalan. Detta tyder på att de matematiskt optimerade trajektorierna förbättrade prestandan, och troligen kan appliceras i många olika fall för att förbättra nedströmsproccesser för biologiska läkemedel. (Less)
Please use this url to cite or link to this publication:
author
Wacker, Philipp LU
supervisor
organization
course
KETM05 20251
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Trajectory optimization, Chromatographic separation, Biopharmaceuticals, Numerical modeling, Downstream processing, Purification
language
English
id
9193662
date added to LUP
2025-06-10 13:11:04
date last changed
2025-06-10 13:11:04
@misc{9193662,
  abstract     = {{In this thesis, the traditional way of conducting chromatographic separation by linear elution was re-evaluated by introducing variable elution buffer trajectories in the mobile phase. Through numerical simulation of ion exchange chromatography using a finite volume method approach, various salt concentration profiles were optimized for improving the separation of components, using yield and productivity as the performance metrics. In three different case studies, the procedures were numerically trialed on peptide, viral genome, and protein mixtures. The protein case results were also verified in a lab-scale experiment, utilizing optimized model control parameters. 

Optimization of the buffer trajectory was studied using two approaches, namely single- and multiple shooting optimization, where single-shooting became the main method applied in this thesis. Due to a high co-variance between variables and the requirement of multiple levels of optimization, the objective functions became noisy and non-convex. This was solved by adjusting the objective of the pooling algorithm and by introducing genetic algorithms, where Differential Evolution was used for singular objectives and the Non-Sorting Genetic Algorithm was used for multiple objective optimization. 

The optimization results varied significantly in different cases, but the step elution (i.e., zero order hold) methods were generally deemed to outperform other hold orders. At higher trajectory degrees of freedom, the optimization convergence worsened significantly. The obtained trajectory solutions showed large performance improvements in the viral vector case, while resulting in minor improvements in the protein case and intermediate enhancement in the peptide case. There was generally no increase in optimal yield, but large increases in performance were often found from the productivity perspective. However, when including the column productivity in the objective, only the first couple of segments were proven to influence the result. This indicates that the total elution time is a critical factor as the end of the elution window often contained a lot of 'dead' time with no performance impact. It was also shown that optimizing an additional flow rate trajectory yielded little improvement compared to only applying a buffer trajectory at a constant flow rate. 

The experimentally trialed protein case was calibrated to the numerical simulation model and showed a large improvement in peak separation. Although the model predicted little performance improvement between a linear gradient and the segmented variable profile, the practical experiment was deemed to perform significantly better, due to an improved peak separation at a similar peak retention time for the target component. Thus, the trajectory optimizations were deemed to be a success. Their implementation in a lab-scale chromatography system was shown to be straight-forward, and could likely lead to an increased production efficiency in many other applications, enhancing the downstream process of biopharmaceuticals.}},
  author       = {{Wacker, Philipp}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Elution Trajectory Optimization for Chromatographic Separation}},
  year         = {{2025}},
}