Integrating text mining and knowledge graph to enhance biopharmaceutical process optimization

Area of Science:

• Biopharmaceutical Process Development
• Glycosylation Profiling
• Computational Biology

Background:

• Consistent therapeutic protein quality requires understanding manufacturing process parameters and glycosylation.
• Cell culture is critical, as glycoprotein structure depends on raw materials, cell genetics, and process controls.
• Existing research on cell culture conditions and glycosylation is fragmented, hindering systematic analysis.

Purpose of the Study:

• To develop a framework for extracting and integrating relationships between cell culture conditions and glycosylation profiles from scientific literature.
• To enable data-driven decision-making and actionable insights for biopharmaceutical process development.
• To overcome fragmentation in published research for a systematic understanding of bioprocess interactions.

Main Methods:

• Developed a text-mining pipeline to extract relationships between cell culture conditions and glycosylation from literature.
• Implemented dictionary-based standardization and ontological classification for enhanced precision.
• Integrated extracted relationships into a knowledge graph for visualizing direct and indirect connections.

Main Results:

• The framework achieved an 88% F1-score in relation extraction.
• Successfully revealed hidden relationships between process parameters and glycan attributes.
• A web-based interface allows dynamic exploration and visualization of bioprocess relationships.

Conclusions:

• The knowledge graph technology and interpretable analytics empower researchers to optimize therapeutic glycan profiles.
• Accelerates manufacturing process development through data-driven bioprocess optimization.
• Represents a significant advancement in understanding and controlling glycosylation in biopharmaceutical manufacturing.