The recombinant adeno-associated virus (rAAV) is a viral vector technology for gene therapy that is considered the safest and most effective way to repair single-gene abnormalities in non-dividing cells. However, improving the viral titer productivity in rAAV production remains challenging. The first step to this end is to effectively monitor the process state variables (cell density, GLC, GLN, LAC, AMM, and rAAV viral titer) to improve the control performance for an enhanced productivity. However, the current approaches to monitoring are expensive, laborious, and time-consuming. This paper presents an extended Kalman filter (EKF) approach used to monitor the rAAV production using the online viable cell density measurements and estimating the other state variables measured at a low frequency. The proposed EKF uses an unstructured mechanistic kinetic model applicable in the upstream process. Three datasets were used for parameter estimation, calibration, and testing, and the data were collected from the production of rAAV through a triple-plasmid transfection of HEK293SF-3F6 cells. Overall, the proposed approach accurately estimated metabolite concentrations and the rAAV production yield. Therefore, the approach has a high potential to be extended to an online soft sensor and to be classified as a cost-effective and fast approach to the monitoring of rAAV production.

Training Deep Learning (DL) models with missing labels is a challenge in diverse engineering applications. Missing value imputation methods have been proposed to try to address this problem, but their performance is affected with Massive Proportion of Missing Labels (MPML). This paper presents a approach for handling MPML in Multivariate Long-Term Time Series Forecasting. It is an two-step process where interpolation (using Gaussian Processes Regression (GPR) and domain knowledge from experts) and prediction model are separated to enable the integration of prior domain knowledge. First, a set of samples of the possible interpolation of the missing outputs are generated by the GPR based on the domain knowledge. Second, the observed input sensor data and interpolated labels from GPR are used to train the prediction model. We evaluated our approach with the development of a soft-sensor with one real datasets to forecast the biomass during recombinant adeno-associated virus (rAAV) production in bioreactors. Our experimental results demonstrate the potential of the approach through quantitative evaluation of the generated forecasts in a case that would be extremely difficult to train a DL model due to MPML.