Artificial Inteligence for monitoring bioprocesses
The CPV of the Future project has evaluated the use of artificial intelligence (AI) as a complement to advanced statistics and multivariate analysis to monitor ongoing bioprocesses. Professor Francisco Valero, leader of the Bioprocess Engineering and Applied Biocatalysis group (ENG4BIO) of the UAB Department of Chemical, Biological and Environmental Engineering, and Toni Manzano, from the company Aizon, specialized in the application of AI solutions in the pharmaceutical industry, summarize the main conclusions of the study.
CPV is the third stage of Process Validation, which is a requirement in the pharmaceutical industry during drug manufacturing. Since Stage 3 is typically a long manufacturing phase, extensive data is accumulated, trended, and analyzed during this stage. Thus, CPV highly recommends process automation, Process Analytical Technologies (PAT) and a deep knowledge of the manufacturing process and drug product in order to interact with the process, avoiding deviations and ensuring the expected product performance and product quality. Continuous variability is part of the reality around manufacturing biological operations and the established conditions (EC), critical quality attributes (CQA), and critical process parameters (CPP) are not enough to describe a real and complete picture of the bioprocess. Statistics and Multivariate Analysis can be complemented with artificial intelligence (AI) to evaluate the on-going process. Under this context, artificial intelligence allows predicting, classifying, recognizing and recommending improvements to the process, which leads to enhanced product quality.
The CPV of the Future project was designed to obtain deep knowledge of the production of recombinant proteins under hypoxic conditions in the cell factory Komogataella phaffii, The applied methodology, the results and conclusions should be generalized across any biomanufacturing operation. Therefore, the identified AI opportunities to mitigate challenges introduced by uncertainty or to augment a continuous multivariate control in real time, provides an added value to the current Statistical Process Control. A set of fed-batches were produced to generate the raw data used to train AI Models which finally controlled the fed-batch process by giving real time feedback to the bioreactor based on the on-going acquired during the process, maintaining the respiratory quotient (RQ) on the set point acting on agitation rate (see figure). The first results obtained has been very promising considering the inherent variability of biological bioprocesses.
The Parenteral Drug Association (PDA) coordinated the CPV of the Future project and the PQRI (USA) and the Science and Innovation Ministry (Spain) partially funded the project. The project was designed to establish a standard procedure for continued process verification (CPV) in bioprocesses applying Artificial Intelligence (AI), Internet of Things (IoT) and cloud technologies as valid mechanisms for the process control in drug manufacturing, The production of recombinant proteins under hypoxic conditions in the cell factory K. phaffii was selected as case study. The group of Bioprocess Engineering and Applied Biocatalisis (ENG4BIO) of the Department of Chemical, Biological and Environmental Engineering of the UAB participates in the project providing the biotech SME, the set of experiments and the lab equipment for batch and fed-batch production. Infors and Bluesense provided the software EVE and the O2 and CO2 gas analyzers sensors connected to the bioreactor and Aizon brought the AI SME, IoT, cloud and AI tools for process control. The rest of the PDA team supported the project from a pharmaceutical perspective.