Developing global regression models for metabolite concentration prediction regardless of cell line

Area of Science:

• Biotechnology and Bioengineering
• Process Analytical Technology (PAT)
• Chemometrics and Spectroscopy

Background:

• Process Analytical Technology (PAT) encourages innovative techniques for real-time process monitoring in drug manufacturing.
• Raman spectroscopy with chemometrics can predict critical parameters in mammalian cell cultures.
• Developing robust models requires extensive batch data, posing a challenge for new cell lines.

Purpose of the Study:

• To develop global regression models applicable across different cell lines, reducing the need for extensive new data.
• To demonstrate the feasibility of such models for both mammalian (CHO, HeLa) and insect (Sf9) cell cultures.
• To evaluate the transferability and predictive accuracy of global models for a new cell line (HEK).

Main Methods:

• Utilized Raman spectroscopy coupled with chemometric tools for in-line, real-time analysis.
• Developed global regression models incorporating data from CHO, HeLa, and Sf9 cell lines.
• Validated the global models by predicting glucose and lactate concentrations in HEK cell cultures.

Main Results:

• Successfully developed and demonstrated regression models for CHO, HeLa, and Sf9 cell lines.
• Global models showed suitable predictions for glucose and lactate in HEK cell cultures.
• Adding a single HEK cell culture batch significantly improved model predictive ability.

Conclusions:

• Global regression models are feasible and effective for monitoring diverse cell cultures.
• This approach substantially reduces the resources and time needed for new cell line model development.
• The study supports the adoption of PAT by enabling more efficient and accurate bioprocess monitoring.