Model-based approach for predicting the impact of genetic modifications on product yield in biopharmaceutical manufacturing—Application to influenza vaccine production

Figures

• 

  Fig 1. Parameter distributions for the kinetic parameter log10(kCSyn) after a bootstrap parameter estimation with n = 1000.

• 

  Fig 2. Virus release dynamics for an infection of 7 ⋅ 105 cells at MOI 1. Experimental data are shown as circles with error bars while the simulated values are illustrated as solid lines for five SGOs and the transduction control.

• 

  Fig 3. Parameter distributions for MGOs, that were investigated experimentally.

• 

  Fig 4. Lumped RMS values for all strategies after comparison of MGO simulations to the experimental data (Fig 5 and Fig H in S1 Text).

• 

  Fig 5. Virus release dynamics for an infection of 7 ⋅ 105 cells at MOI 10−4 for the additive and the shift strategy and all MGOs, that were investigated experimentally. Experimental data are shown as circles with error bars while the simulated values are illustrated as green dashed and blue dotted lines.

• 

  Fig 6. Virus yields after 72 hours of the best three and the least promising combinations using the constructed MGO parameter distributions consisting of two to five SGOs. The coloring of the bars classifies the number of genes modified in the engineered cell lines. The darker the color of the bars the more gene modifications were considered. The order of the gene modifications on the y-axis from left to right corresponds to the order of the stepwise lentiviral transductions made to generate the MGOs. The parental cell line A549 and the best SGO (gene modification in NXF1) were illustrated as dashed blue and solid yellow vertical lines, respectively.

• 

  Fig 7. General workflow for the application of the presented methodology on a broad range of biopharmaceutical manufacturing processes.

• 

  Fig 8. Mathematical model with different scales based on [14, 16]. Production of virus particles in mammalian cells can be investigated in T-flask experiments (A). The extracellular model is represented by the dashed box (B). Uninfected target cells can be infected by virus particles. Both infected and uninfected cells can become apoptotic. The intracellular model describes the replication of the viral genome (C). The mathematical description of the intracellular model starts with the virus in the endosome. Different intracellular stages can be influenced by several HCFs. After transcription and translation of the viral genome new virus particles can be released to the surrounding medium. The rates affected by genetic modification (arrows in C) are marked with the yellow asterisks. As an example, the distribution of the matrix protein M1 (PM1) and the parameter kM1Bind (red arrow in C) are shown for two different time points (D).

• 

  Fig 9. Distribution of the parameter kBindM1 in MGO 3 (blue rectangles) for the low impact (A), high impact (B), additive (C) and shift strategy (D). For the low and high impact strategy the distributions were assembled with five Gaussians (red curves). The position of the Gaussians were obtained as described in the corresponding subsections. For the additive and shift strategy, bootstrap parameter estimates (solid and dashed red curves) were used to generate the distribution of the MGO (blue rectangles). Distributions are normalized with respect to the total number of infected cells.