A novel data gathering approach that relies on text mining can significantly enhance the understanding of the complex relationships between culture conditions and glycosylation, according to recent research led by Chuming Chen, PhD, at the Delaware Biotechnology Institute. This method aims to bridge the gap in knowledge that has persisted despite extensive experimentation in biopharma.
The glycosylation profile of a protein, which is crucial for its therapeutic efficacy, is influenced by various culture conditions. While biopharma has excelled in experimental approaches to assess these impacts, the industry has struggled to create generalized models that connect culture conditions to glycosylation outcomes. Chen notes that the fragmentation of existing knowledge stems from variability in study conditions, complicating the establishment of clear relationships.
To address this challenge, Chen and his collaborators developed an automated text mining pipeline that extracts relationships from unstructured scientific literature with an impressive 88% accuracy. This innovation not only streamlines data collection but also leads to the creation of a Bioprocess Knowledge Graph Database, which captures both direct and indirect associations between process parameters and glycan outcomes. The resulting web interface allows researchers to dynamically explore these complex relationships, facilitating more informed decision-making in therapeutic protein manufacturing. As the project evolves, the team plans to integrate deep learning and large language models to further enhance the system’s capabilities.
Open the full market picture for your next decision →
