Cross-flow ultrafiltration is a powerful tool used in bioprocesses to concentrate and separate biopharmaceuticals. However, there is a major challenge for its effective use: the fouling rate and decrease in the permeate flux vary substantially depending on the concentration and characteristics of the incoming feed, which causes variations in the process durations. We developed a hybrid model for use in predicting the flux evolution and duration of cross-flow ultrafiltration processes for various proteins, membrane types, input parameters, and filtration modes. The trained hybrid model is able to determine the process duration with an average normalized root-mean-square error of less than 2.5. It has superior adaptation to varying filtration characteristics compared to the mechanistic film theory model and can handle both batch- and fed-batch ultrafiltration using the same training set. In addition, the presented hybrid model can be used as a digital twin to simulate virtual processes with varying input parameters and different batch modes, which is a valuable tool for efficient process development or optimization.

Ultrafiltration is a powerful method used in virtually every pharmaceutical bioprocess. Depending on the process stage, the product-to-impurity ratio differs. The impact of impurities on the process depends on various factors. Solely mechanistic models are currently not sufficient to entirely describe these complex interactions. We have established two hybrid models for predicting the flux evolution, the protein rejection factor and two components’ concentration during crossflow ultrafiltration. The hybrid models were compared to the standard mechanistic modeling approach based on the stagnant film theory. The hybrid models accurately predicted the flux and concentration over a wide range of process parameters and product-to-impurity ratios based on a minimum set of training experiments. Incorporating both components into the modeling approach was essential to yielding precise results. The stagnant film model exhibited larger errors and no predictions regarding the impurity could be made, since it is based on the main product only. Further, the developed hybrid models exhibit excellent interpolation properties and enable both multi-step ahead flux predictions as well as time-resolved impurity forecasts, which is considered to be a critical quality attribute in many bioprocesses. Therefore, the developed hybrid models present the basis for next generation bioprocessing when implemented as soft sensors for real-time monitoring of processes.